1 2018-11-09 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/87621
4 * tree-vect-loop.c (vectorizable_reduction): Handle reduction
5 op with only phi inputs.
6 * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
7 (ch_base::copy_headers): Run CSE on copied loop headers.
8 (pass_ch_vect::process_loop_p): Simplify.
10 2018-11-09 Alexandre Oliva <oliva@adacore.com>
12 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
13 for non-w64 x86_64 biarch.
15 2018-11-09 Alexandre Oliva <aoliva@redhat.com>
17 PR rtl-optimization/86438
18 * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
19 of in_b for the compare if in_b is SET_DEST.
22 * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
25 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
27 * tree-vrp.c (value_range::check): Do not access internals
29 (value_range::singleton_p): Same.
30 (value_range::type): Same.
31 (vrp_finalize): Use value_range API.
33 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
35 * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
37 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
39 * vr-values.c (vr_values::get_value_range): Use value_range API
40 instead of piecing together ranges.
41 (vr_values::update_value_range): Same.
43 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
45 * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
46 of performing ad-hoc calculations.
47 * tree-ssanames.c (set_range_info): New overloaded function
48 accepting value_range &.
49 (get_range_info): Same.
50 * tree-ssanames.h (set_range_info_raw): Remove.
51 (set_range_info): New prototype.
52 (get_range_info): Same.
53 * tree-vrp.h (value_range::null_p): Rename to zero_p.
54 * tree-vrp.c (value_range::null_p): Same.
56 2018-11-09 Jan Hubicka <jh@suse.cz>
58 * tree.c (fld_type_variant_equal_p): Test user align flag.
59 (flt_type_variant): Copy user align flag.
60 (fld_incomplete_type_of): Clear it.
62 2018-11-09 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
64 * config/arm/neon.md (div<mode>3): New pattern.
66 2018-11-08 Andi Kleen <ak@linux.intel.com>
68 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
69 (OPTION_MASK_ISA_PTWRITE_UNSET): New.
70 (ix86_handle_option): Handle OPT_mptwrite.
71 * config/i386/cpuid.h (bit_PTWRITE): Add.
72 * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
73 * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
74 * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
75 * config/i386/i386.c (ix86_target_string): Handle ptwrite.
76 (ix86_option_override_internal): Handle PTA_PTWRITE.
77 (ix86_valid_target_attribute_inner_p): Define ptwrite.
78 (def_builtin2): Force UINT64 to be 64bit only.
79 * config/i386/i386.h (TARGET_PTWRITE): Add.
80 (TARGET_PTWRITE_P): Add.
82 * config/i386/i386.md: Define ptwrite.
83 * config/i386/i386.opt: Add -mptwrite.
84 * config/i386/immintrin.h (_ptwrite64): Add.
86 * doc/extend.texi: Document __builtin_ia32_ptwrite*.
87 * doc/invoke.texi: Document -mptwrite.
89 2018-11-08 Peter Bergner <bergner@linux.ibm.com>
91 PR rtl-optimization/87600
92 * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
93 * lra-constraints.c (process_alt_operands): Skip illegal hard
94 register usage. Prefer reloading non hard register operands.
97 PR rtl-optimization/87600
98 * gcc.dg/pr87600.h: New file.
99 * gcc.dg/pr87600-1.c: New test.
100 * gcc.dg/pr87600-2.c: Likewise.
102 2018-11-08 Sandra Loosemore <sandra@codesourcery.com>
105 * doc/invoke.texi (Optimize Options): Clarify default behavior
106 for -fno-sched-interblock and -fno-sched-spec.
108 2018-11-08 Roman Geissler <roman.geissler@amadeus.com>
110 * collect2.c (linker_select): Add USE_LLD_LD.
111 (ld_suffixes): Add ld.lld.
112 (main): Handle -fuse-ld=lld.
113 * common.opt (-fuse-ld=lld): New option.
114 * doc/invoke.texi (-fuse-ld=lld): Document.
115 * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
117 2018-11-08 Paul Koning <ni1d@arrl.net>
119 * config/pdp11/constraints.md: Add "Z" series constraints for use
120 with pre-dec and post-inc addressing.
121 * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
122 (pdp11_expand_operands): Add int argument (word count).
123 (pdp11_sp_frame_offset): Delete.
124 (pdp11_cmp_length): New function.
125 (pushpop_regeq): New function.
126 * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
128 (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
130 (pdp11_initial_elimination_offset): Ditto.
131 (pdp11_expand_operands): Add word count argument. Bugfixes.
132 (output_move_multiple): Change how pointer adjustment is done.
133 (pdp11_gen_int_label): Correct format.
134 (output_ascii): Ditto.
135 (pdp11_asm_output_var): Add code for DEC assembler case.
136 (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
138 (legitimate_const_double_p): Ditto.
139 (pdp11_register_move_cost): Adjust for new register classes.
140 (pdp11_regno_reg_class): Ditto.
141 (expand_block_move): Delete.
142 (pushpop_regeq): New function.
143 (pdp11_legitimate_address_p): Bugfix in check for constant
145 (pdp11_sp_frame_offset): Delete.
146 (pdp11_reg_save_size): New helper function for new frame layout.
147 (output_addr_const_pdp11): Remove CONST_DOUBLE case.
148 (pdp11_expand_shift): Bugfix in check for constant shift count.
149 (pdp11_shift_length): Ditto.
150 (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
151 (pdp11_cmp_length): New function.
152 * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
153 some compile options.
154 (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
155 (CALL_USED_REGISTERS): Ditto.
156 (ELIMINABLE_REGS): Ditto.
157 (REGISTER_NAMES): Ditto.
158 (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
160 (REG_CLASS_NAMES): Ditto.
161 (REG_CLASS_CONTENTS): Ditto. Also remove
162 HARD_FRAME_POINTER_REGNUM.
163 (CPU_REG_CLASS): New macro.
164 (CLASS_MAX_NREGS): Adjust for new register classes.
165 (FUNCTION_PROFILER): Make no-op.
166 (may_call_alloca): Remove unused declaration.
167 (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
168 (ASM_OUTPUT_SKIP): Fix format.
169 * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
170 (HARD_FRAME_POINTER_REGNUM): Remove.
172 (*rts): Rename. Remove epilogue related checks.
173 (cmpsi, cmpdi): New insn.
174 (cbranch<mode>4): Change to apply to SI and DI modes as well.
175 (mov<mode>): Change constraints to enforce that push/pop
176 destination cannot use the same register as source.
177 (*mov<mode><cc_cc>): Ditto.
178 (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
179 at assembly output rather than as RTL expander.
180 (zero_extendqihi2): Bugfix in check for same registers.
181 (adddi3_nocc): Bugfix in check for constant operand.
182 (addsi3_nocc): Ditto.
183 (subdi3_nocc): Ditto.
184 (subsi3_nocc): Ditto.
185 (negdi2_nocc): Copy input to pdp11_expand_operands.
186 (negsi2_nocc): Ditto.
187 (bswap2_nocc): Ditto.
188 * config/pdp11/pdp11.opt (mlra): Fix documentation.
189 * config/pdp11/t-pdp11: Use -Os.
191 2018-11-08 Richard Earnshaw <rearnsha@arm.com>
193 * config/arm/parsecpu.awk (/alias/): New parsing rule.
194 (/begin cpu/): Check that the cpu name hasn't been previously defined.
195 (gen_comm_data): Print out CPU alias tables.
196 (check_cpu): Match aliases when checking the CPU name.
197 * config/arm/arm-protos.h (cpu_alias): New structure.
198 (cpu_option): Add entry for aliases.
199 * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
200 strongarm1100 and strongarm1110.
201 (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
202 (config/arm/arm-generic.md): Remove redundant references to
203 strongarm110, strongarm1100 and strongarm1110.
204 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
205 Scan aliases for additional hints.
206 (arm_parse_cpu_option_name): Also match a cpu name against the list
208 * config/arm/arm-tables.opt: Regenerated.
209 * config/arm/arm-tune.md: Regenerated.
211 2018-11-08 Jakub Jelinek <jakub@redhat.com>
213 * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
214 BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
215 BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
216 BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
217 BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
218 BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
219 * gengtype.c (open_base_files): Add omp-general.h.
220 * gimple.c (gimple_build_omp_critical):
221 (gimple_build_omp_taskgroup): Add CLAUSES argument. Call
222 gimple_omp_taskgroup_set_clauses.
223 (gimple_build_omp_atomic_load): Add mo argument, call
224 gimple_omp_atomic_set_memory_order.
225 (gimple_build_omp_atomic_store): Likewise.
226 (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
227 * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
229 (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
230 of GSS_OMP_SINGLE_LAYOUT, adjust comments.
231 * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
232 and GF_OMP_ATOMIC_MEMORY_ORDER. Remove GF_OMP_ATOMIC_SEQ_CST, use
233 different value for GF_OMP_ATOMIC_NEED_VALUE.
234 (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
236 (struct gimple_statement_omp_single_layout): And remove here.
237 (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
238 than gimple_statement_omp_single_layout.
239 (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
241 (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
242 (gimple_omp_subcode): Formatting fix.
243 (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
244 gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
245 gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
246 gimple_omp_teams_host, gimple_omp_teams_set_host,
247 gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
248 gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
249 gimple_omp_taskgroup_set_clauses): New inline functions.
250 (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
251 (gimple_build_omp_atomic_store): Likewise.
252 (gimple_omp_atomic_seq_cst_p): Remove.
253 (gimple_omp_atomic_memory_order): New function.
254 (gimple_omp_atomic_set_seq_cst): Remove.
255 (gimple_omp_atomic_set_memory_order): New function.
256 (gimple_build_omp_taskgroup): Add clauses argument.
257 * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
258 (dump_gimple_omp_task): Print taskwait with depend clauses.
259 (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
260 dump_omp_atomic_memory_order.
261 (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
262 * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
263 GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
264 (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
265 renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
266 ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
267 ORT_UNTIED_TASKLOOP enumerators.
268 (enum gimplify_defaultmap_kind): New.
269 (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
270 target_map_pointers_as_0len_arrays members, add defaultmap.
271 (new_omp_context): Initialize defaultmap member.
272 (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
273 (maybe_fold_stmt): Don't fold even in host teams regions.
274 (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
275 ORT_WORKSHARE. Test ctx->defaultmap[GDMK_SCALAR] instead of
276 ctx->omp_firstprivatize_variable.
277 (omp_add_variable): Don't add private/firstprivate for VLAs in
279 (omp_default_clause): Print "taskloop" rather than "task" if
281 (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
282 Handle new defaultmap clause kinds.
283 (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE. Allow simd
284 iterator to be lastprivate or private. Fix up diagnostics if linear
285 is used on collapse>1 simd iterator.
286 (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
287 (gimplify_omp_depend): New function.
288 (gimplify_scan_omp_clauses): Add shared clause on parallel for
289 combined parallel master taskloop{, simd} if taskloop has
290 firstprivate, lastprivate or reduction clause. Handle
291 OMP_CLAUSE_REDUCTION_TASK diagnostics. Adjust tests for
292 ORT_COMBINED_TEAMS. Gimplify depend clauses with iterators. Handle
293 cancel and simd OMP_CLAUSE_IF_MODIFIERs. Handle
294 OMP_CLAUSE_NONTEMPORAL. Handle new defaultmap clause kinds. Handle
295 OMP_CLAUSE_{TASK,IN}_REDUCTION. Diagnose invalid conditional
297 (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL. Handle
298 GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.
299 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL. Handle
300 OMP_CLAUSE_{TASK,IN}_REDUCTION.
301 (gimplify_omp_task): Handle taskwait with depend clauses.
302 (gimplify_omp_for): Add shared clause on parallel for combined
303 parallel master taskloop{, simd} if taskloop has firstprivate,
304 lastprivate or reduction clause. Use ORT_TASKLOOP or
305 ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK. Adjust
306 tests for ORT_COMBINED_TEAMS. Handle C++ range for loops with
307 NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS. Firstprivatize
308 __for_end and __for_range temporaries on OMP_PARALLEL for
309 distribute parallel for{, simd}. Move OMP_CLAUSE_REDUCTION
310 and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
311 sandwiched in between two taskloops.
312 (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
313 instead of ctx->omp_firstprivatize_variable.
314 (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
315 ORT_COMBINED_HOST_TEAMS if not inside of target construct. If
316 host teams, use gimplify_and_return_first etc. for body like
317 for target or target data constructs, and at the end call
318 gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
319 (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
320 of OMP_ATOMIC_SEQ_CST, pass it as new argument to
321 gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
322 gimple_omp_atomic_set_seq_cst calls.
323 (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
324 case, handle taskgroup clauses.
325 * lto-streamer-out.c (hash_tree): Handle
326 OMP_CLAUSE_{TASK,IN}_REDUCTION.
327 * Makefile.in (GTFILES): Add omp-general.h.
328 * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
329 BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
330 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
331 BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
332 BUILT_IN_GOMP_LOOP_DOACROSS_START,
333 BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
334 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
335 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
336 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
337 BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
338 BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
339 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
340 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
341 BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
342 BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
343 BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
344 BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
345 BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
346 BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
347 BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
348 * omp-expand.c (workshare_safe_to_combine_p): Return false for
349 non-worksharing loops.
350 (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
351 (determine_parallel_type): Don't combine parallel with worksharing
352 which has _reductemp_ clause.
353 (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
354 GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
355 if there is nonmonotonic modifier or if there is no modifier and no
356 ordered clause. For dynamic and guided schedule without monotonic
357 and nonmonotonic modifier, default to nonmonotonic.
358 (expand_omp_for): Likewise. Adjust expand_omp_for_generic caller, use
359 GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
361 (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
362 there are any reduction clauses.
363 (expand_taskwait_call): New function.
364 (expand_teams_call): New function.
365 (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
366 expand_teams_call for it. Formatting fix. Handle taskwait with
368 (expand_omp_for_generic): Add SCHED_ARG argument. Handle expansion
369 of worksharing loops with task reductions.
370 (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
371 expansion of worksharing loops with task reductions.
372 (expand_omp_sections): Handle expansion of sections with task
374 (expand_omp_synch): For host teams call expand_omp_taskreg.
375 (omp_memory_order_to_memmodel): New function.
376 (expand_omp_atomic_load, expand_omp_atomic_store,
377 expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
378 instead of gimple_omp_atomic_seq_cst_p.
379 (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
380 depend clauses as a standalone directive.
381 * omp-general.c (enum omp_requires): New variable.
382 (omp_extract_for_data): Initialize have_reductemp member. Allow
383 NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
384 GT_EXPR loops depending on incr sign. Formatting fixes.
385 * omp-general.h (struct omp_for_data): Add have_reductemp member.
386 (enum omp_requires): New enum.
387 (omp_requires_mask): Declare.
388 * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
390 * omp-low.c (struct omp_context): Add task_reductions and
391 task_reduction_map fields.
392 (is_host_teams_ctx): New function.
393 (is_taskreg_ctx): Return true also if is_host_teams_ctx.
394 (use_pointer_for_field): Use is_global_var instead of
395 TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
397 (build_outer_var_ref): Ignore taskgroup outer contexts.
398 (delete_omp_context): Release task_reductions and task_reduction_map.
399 (scan_sharing_clauses): Don't add any fields for reduction clause on
400 taskloop. Handle OMP_CLAUSE__REDUCTEMP_. Handle
401 OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
402 modifier. Don't ignore shared clauses in is_host_teams_ctx contexts.
403 Handle OMP_CLAUSE_NONTEMPORAL.
404 (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
406 (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
407 clauses with task modifier.
408 (scan_omp_task): Handle taskwait with depend clauses.
409 (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
410 first. Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
411 Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
412 (scan_omp_for): Fix comment formatting.
413 (scan_omp_teams): Handle host teams constructs.
414 (check_omp_nesting_restrictions): Allow teams with no outer
415 OpenMP context. Adjust diagnostics for teams strictly nested into
416 some explicit OpenMP construct other than target. Allow OpenMP atomics
417 inside of simd regions.
418 (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
419 (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
420 taskreg_nesting_level while scanning host teams construct.
421 (task_reduction_read): New function.
422 (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
423 construct. Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
424 clauses. Handle OMP_CLAUSE_REDUCTION with task modifier. Remove
425 second argument create_tmp_var if it is NULL. Don't ignore shared
426 clauses in is_host_teams_ctx contexts. Handle
427 OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
429 (lower_reduction_clauses): Ignore reduction clauses with task
430 modifier. Remove second argument create_tmp_var if it is NULL.
431 Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
432 (lower_send_clauses): Ignore reduction clauses with task modifier.
433 Handle OMP_CLAUSE__REDUCTEMP_. Don't send anything for
434 OMP_CLAUSE_REDUCTION on taskloop. Handle OMP_CLAUSE_IN_REDUCTION.
435 (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
436 rely that it is the last stmt in body so far. Ignore outer taskgroup
438 (omp_task_reductions_find_first, omp_task_reduction_iterate,
439 lower_omp_task_reductions): New functions.
440 (lower_omp_sections): Handle reduction clauses with taskgroup
441 modifiers. Adjust maybe_add_implicit_barrier_cancel caller.
442 (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
443 (lower_omp_for): Likewise. Handle reduction clauses with taskgroup
445 (lower_omp_taskgroup): Handle taskgroup reductions.
446 (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
447 Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
448 (lower_depend_clauses): If there are any
449 OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
450 depend clauses, use a new array format. If OMP_CLAUSE_DEPEND_LAST is
451 seen, assume lowering is done already and return early. Set kind
452 on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
453 (lower_omp_taskreg): Handle reduction clauses with task modifier on
454 parallel construct. Handle reduction clause on taskloop construct.
455 Handle taskwait with depend clauses.
456 (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
457 for host teams constructs.
458 * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
459 nontemporal and _reductemp_ clause entries.
460 (omp_clause_code_name): Likewise.
461 (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
462 OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
463 * tree-core.h (enum omp_clause_code): Add
464 OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
465 (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
466 (struct tree_base): Add omp_atomic_memory_order field into union.
467 Remove OMP_ATOMIC_SEQ_CST comment.
468 (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
469 and OMP_CLAUSE_DEPEND_DEPOBJ.
470 (struct tree_omp_clause): Add subcode.defaultmap_kind.
471 * tree.def (OMP_TASKGROUP): Add another operand, move next to other
472 OpenMP constructs with body and clauses operands.
473 * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
474 (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
475 (OMP_TASKGROUP_CLAUSES): Define.
476 (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
477 OMP_CLAUSE__LOOPTEMP_.
478 (OMP_ATOMIC_SEQ_CST): Remove.
479 (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
480 OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
481 (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
482 OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
483 OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
484 OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
485 OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
486 (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
487 OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
488 OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
490 * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
491 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
492 OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
493 (convert_local_omp_clauses): Likewise. Remove useless test.
494 * tree-parloops.c (create_call_for_reduction_1): Pass
495 OMP_MEMORY_ORDER_RELAXED as new argument to
496 dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
497 * tree-pretty-print.c (dump_omp_iterators): New function.
498 (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
499 OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION. Print
500 reduction modifiers. Handle OMP_CLAUSE_DEPEND_DEPOBJ and
501 OMP_CLAUSE_DEPEND_MUTEXINOUTSET. Print iterators in depend clauses.
502 Print __internal__ for OMP_CLAUSE_DEPEND_LAST. Handle cancel and
503 simd OMP_CLAUSE_IF_MODIFIERs. Handle new kinds of
504 OMP_CLAUSE_DEFAULTMAP. Print conditional: for
505 OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
506 (dump_omp_atomic_memory_order): New function.
507 (dump_generic_node): Use it. Print taskgroup clauses. Print
508 taskwait with depend clauses.
509 * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
510 * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
511 Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
512 * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
513 write_ts_omp_clause_tree_pointers): Likewise.
515 2018-11-08 David Malcolm <dmalcolm@redhat.com>
518 * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
519 "-missed", "-note", and "-all" sub-options.
520 * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
522 (can_early_inline_edge_p): Likewise.
523 (want_early_inline_function_p): Likewise.
524 (want_inline_self_recursive_call_p): Likewise.
525 (recursive_inlining): Likewise.
526 (inline_small_functions): Likewise.
527 (flatten_function): Likewise.
528 (ipa_inline): Likewise.
529 (inline_always_inline_functions): Likewise.
530 (early_inline_small_functions): Likewise.
531 (early_inliner): Likewise.
532 * tree-inline.c (expand_call_inline): Likewise.
534 2018-11-08 David Malcolm <dmalcolm@redhat.com>
536 * pretty-print.c (pp_format): Handle %f.
537 (selftest::test_pp_format): Add test of %f.
538 * pretty-print.h (pp_double): New macro.
540 2018-11-08 David Malcolm <dmalcolm@redhat.com>
542 * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
543 * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
544 (dump_pretty_printer::decode_format): Implement "%C" for
546 (selftest::test_capture_of_dump_calls): Rename "where" to
547 "stmt_loc". Convert test_decl to a function decl and set its
548 location. Add a symbol_table_test RAII instance and a
549 cgraph_node, using it to test "%C" and dump_symtab_node.
551 2018-11-08 Eric Botcazou <ebotcazou@adacore.com>
554 * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
556 2018-11-08 David Malcolm <dmalcolm@redhat.com>
558 * cgraph.c: Include "selftest.h".
559 (saved_symtab): New variable.
560 (selftest::symbol_table_test::symbol_table_test): New ctor.
561 (selftest::symbol_table_test::~symbol_table_test): New dtor.
562 (selftest::test_symbol_table_test): New test.
563 (selftest::cgraph_c_tests): New.
564 * cgraph.h (saved_symtab): New decl.
565 (selftest::symbol_table_test): New class.
566 * selftest-run-tests.c (selftest::run_tests): Call
567 selftest::cgraph_c_tests.
568 * selftest.h (selftest::cgraph_c_tests): New decl.
570 2018-11-08 Richard Biener <rguenther@suse.de>
572 * tree-data-ref.h (lambda_int): New typedef.
573 (lambda_vector_gcd): Adjust.
574 (lambda_vector_new): Likewise.
575 (lambda_matrix_new): Likewise.
576 * tree-data-ref.c (print_lambda_vector): Adjust.
578 2018-11-08 Richard Biener <rguenther@suse.de>
580 PR tree-optimization/87929
581 * tree-complex.c (expand_complex_comparison): Clean EH.
583 2018-11-08 Martin Liska <mliska@suse.cz>
585 * doc/extend.texi: Reword.
586 * predict.c (expr_expected_value_1): Likewise.
588 2018-11-08 Richard Biener <rguenther@suse.de>
590 PR tree-optimization/87913
591 * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
592 of extreme values to ordered comparisons.
594 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
597 * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
599 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
601 * doc/invoke.texi: Remove leading dash from @opindex entries
604 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
607 * doc/invoke.texi (Option Summary): Add -e and --entry.
608 (Link Options): Likewise.
610 2018-11-07 Nathan Sidwell <nathan@acm.org>
613 * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
614 --disable-checking bootstrap.
616 2018-11-07 Hafiz Abid Qadeer <abidh@codesourcery.com>
618 * configure: Regenerated.
620 2018-11-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
623 * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
624 to the mode of the widest field iff the widest field has mode class
625 MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
628 2018-11-07 Nikolai Merinov <n.merinov@inango-systems.com>
630 * common.opt: Add -Wattribute-warning.
631 * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
632 * expr.c (expand_expr_real_1): Add new attribute to warning_at
633 call to allow user configure behavior of "warning" attribute.
635 2018-11-07 Segher Boessenkool <segher@kernel.crashing.org>
637 * target.def: Put @: after every vs., e.g., and i.e. where it is
638 followed by whitespace.
639 * doc/extend.texi: Ditto.
640 * doc/fragments.texi: Ditto.
641 * doc/gimple.texi: Ditto.
642 * doc/implement-c.texi: Ditto.
643 * doc/install.texi: Ditto.
644 * doc/invoke.texi: Ditto.
645 * doc/md.texi: Ditto.
646 * doc/plugins.texi: Ditto.
647 * doc/rtl.texi: Ditto.
648 * doc/sourcebuild.texi: Ditto.
649 * doc/tm.texi.in: Ditto.
650 * doc/ux.texi: Ditto.
651 * doc/tm.texi: Regenerate.
653 2018-11-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
655 * config/arm/arm-cpus.in (ares): New entry.
656 * config/arm/arm-tables.opt: Regenerate.
657 * config/arm/arm-tune.md: Likewise.
658 * doc/invoke.texi (ARM Options): Document ares.
660 2018-11-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
662 * config/aarch64/aarch64-cores.def (ares): Define.
663 * config/aarch64/aarch64-tune.md: Regenerate.
664 * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
666 2018-11-07 Jan Hubicka <jh@suse.cz>
668 * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
669 (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
670 functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
673 2018-11-07 Richard Biener <rguenther@suse.de>
675 PR tree-optimization/87914
676 * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
678 (vectorizable_reduction): Handle shifts and rotates by dispatching
679 to vectorizable_shift.
680 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
681 in-loop uses of vect_nested_cycle defs. Merge cycle and internal
683 (vectorizable_shift): Export and handle being called as
685 (vect_analyze_stmt): Call vectorizable_shift after
686 vectorizable_reduction.
687 * tree-vectorizer.h (vectorizable_shift): Declare.
689 2018-11-07 Jan Hubicka <jh@suse.cz>
691 * ipa-devirt.c (odr_types_equivalent_p): Expect constants
692 than const decls in TREE_VALUE of enum.
693 (dump_type_inheritance_graph): Improve duplicate dumping.
694 (free_enum_values): New.
695 (build_type_inheritance_graph): Use it.
696 * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
697 which are not main variants or not ODR types.
698 (verify_type_variant): Expect variants to have no TYPE_VALUES.
700 2018-11-07 Richard Biener <rguenther@suse.de>
702 * ipa-inline.c (want_inline_small_function_p): Compute
703 big_speedup_p lazily and last.
705 2018-11-07 Jan Hubicka <jh@suse.cz>
707 * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
708 building incomplete variant of complete type.
709 (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
710 variant of complete type.
712 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
714 * config/mips/mips.c: Fix typo in documentation of
715 mips_loongson_ext2_prefetch_cookie.
716 (mips_option_override): fix brain twister logical.
717 * config/mips/mips.h: Fix typo in documentation of
718 ISA_HAS_CTZ_CTO and define pattern.
719 * config/mips/mips.md (prefetch): Hoist EXT2 above
721 (prefetch_indexed): Hoist EXT2 above the EXT block.
723 2018-11-07 Jan Hubicka <jh@suse.cz>
725 * tree.c (free_lang_data_in_type): Add fld parameter; simplify
726 return and parameter types of function and method types.
727 (free_lang_data_in_cgraph): Update.
729 2018-11-07 Martin Liska <mliska@suse.cz>
731 PR rtl-optimization/87868
732 * postreload-gcse.c (eliminate_partially_redundant_load): Set
733 threshold to max_count if we would overflow.
734 * profile-count.h: Make max_count a public constant.
736 2018-11-07 Martin Liska <mliska@suse.cz>
738 * mem-stats.h: Fix GNU coding style.
740 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
742 * config/mips/gs264e.md: New.
743 * config/mips/mips-cpus.def: Define gs264e.
744 * config/mips/mips-tables.opt: Regenerate.
745 * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
747 (mips_issue_rate): Add support for gs264e.
748 (mips_multipass_dfa_lookahead): Likewise.
749 * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
750 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
751 (MIPS_ASE_MSA_SPEC): New.
752 (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
753 (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
754 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
755 * config/mips/mips.md: Include gs264e.md.
756 (processor): Add gs264e.
757 * config/mips/mips.opt (MSA): Use Mask instead of Var.
758 * doc/invoke.texi: Add gs264e to supported architectures.
760 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
762 * config/mips/gs464e.md: New.
763 * config/mips/mips-cpus.def: Define gs464e.
764 * config/mips/mips-tables.opt: Regenerate.
765 * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
767 (mips_issue_rate): Add support for gs464e.
768 (mips_multipass_dfa_lookahead): Likewise.
769 (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
770 * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
771 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
772 (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
773 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
774 * config/mips/mips.md: Include gs464e.md.
775 (processor): Add gs464e.
776 * doc/invoke.texi: Add gs464e to supported architectures.
778 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
780 * config/mips/loongson3a.md: Rename to ...
781 * config/mips/gs464.md: ... here.
782 * config/mips/mips-cpus.def: Define gs464; Add loongson3a
783 as an alias of gs464 processor.
784 * config/mips/mips-tables.opt: Regenerate.
785 * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
786 instead of PROCESSOR_LOONGSON_3A.
787 (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
789 (mips_option_override): Enable MMI and EXT for gs464.
790 * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
791 Rename TUNE_LOONGSON_3A to TUNE_GS464.
792 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
793 (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
794 ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
796 * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
797 (processor): Add gs464;
798 * doc/invoke.texi: Add gs464 to supported architectures.
800 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
802 * config/mips/mips-protos.h
803 (mips_loongson_ext2_prefetch_cookie): New prototype.
804 * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
805 (mips_option_override): Enable TARGET_LOONGSON_EXT when
806 TARGET_LOONGSON_EXT2 is true.
807 * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
808 __mips_loongson_ext2, __mips_loongson_ext_rev=2.
809 (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
810 (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
811 TARGET_LOONGSON_EXT2.
812 (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
813 (define_insn "ctz<mode>2"): New insn pattern.
814 (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
815 (define_insn "prefetch_indexed_<mode>"): Include
816 TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
817 * config/mips/mips.opt (-mloongson-ext2): Add option.
818 * gcc/doc/invoke.texi (-mloongson-ext2): Document.
820 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
822 * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
824 (MIPS_ASE_LOONGSON_EXT_SPEC): New.
825 (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
827 (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
828 * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
829 <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
830 instead of TARGET_LOONGSON_3A.
831 * config/mips/mips.opt (-mloongson-ext): Add option.
832 * gcc/doc/invoke.texi (-mloongson-ext): Document.
834 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
836 * config.gcc (extra_headers): Add loongson-mmiintrin.h.
837 * config/mips/loongson.md: Move to ...
838 * config/mips/loongson-mmi.md: here; Adjustment.
839 * config/mips/loongson.h: Move to ...
840 State as deprecated. Include loongson-mmiintrin.h for back
841 compatibility and warning.
842 * config/mips/loongson-mmiintrin.h: ... here.
843 * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
844 mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
845 TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
846 (mips_option_override): Make sure MMI use hard float;
847 (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
848 mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
849 mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
850 TARGET_LOONGSON_VECTORS.
851 * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
852 (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
853 (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
854 (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
856 (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
857 TARGET_LOONGSON_VECTORS.
858 * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
859 TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
860 (Loongson MMI patterns): Include loongson-mmi.md instead of
862 * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
863 * gcc/doc/invoke.texi (-mloongson-mmi): Document.
865 2018-11-07 Richard Biener <rguenther@suse.de>
868 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
869 BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
871 2018-11-07 Alexandre Oliva <aoliva@redhat.com>
873 PR rtl-optimization/87874
874 * lra.c (lra_substitute_pseudo): Do not create a subreg for
877 2018-11-06 Aaron Sawdey <acsawdey@linux.ibm.com>
879 * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
880 if not in indexed or indirect form.
881 (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
882 (bswap<mode>2_store): Ditto.
884 2018-11-06 Richard Earnshaw <rearnsha@arm.com>
886 * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
889 2018-11-06 Richard Biener <rguenther@suse.de>
891 PR tree-optimization/86850
892 * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
893 instead of src.m_vec.
895 2018-11-06 Jan Hubicka <jh@suse.cz>
897 * tree.c (fld_simplified_type_name): Break out form ...
898 (free_lang_data_in_type): ... here.
899 (fld_type_variant_equal_p): Use it.
901 2018-11-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
903 * config/default-d.c: Include memmodel.h.
905 * config/sol2-d.c: New file.
906 * config/t-sol2 (sol2-d.o): New rule.
907 * config.gcc <*-*-solaris2*>: Set d_target_objs,
910 2018-11-06 Jan Hubicka <jh@suse.cz>
912 * tree.c (fld_type_variant): Also copy alignment; be sure that
913 new variant is equal.
915 2018-11-06 Ilya Leoshkevich <iii@linux.ibm.com>
918 * config/s390/s390.md: Add relative_long attribute.
920 2018-11-06 Jan Hubicka <jh@suse.cz>
922 * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
923 * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
924 * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
925 stream TYPE_NEEDS_CONSTRUCTING.
926 * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
927 * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
929 2018-11-06 Richard Biener <rguenther@suse.de>
931 * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
933 (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
935 2018-11-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
937 * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
938 REG_CLASS_CONTENTS[GEN_REGS].
939 (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
941 2018-11-06 Jan Hubicka <jh@suse.cz>
943 * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
945 2018-11-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
948 * config/sol2.h (ASAN_CC1_SPEC): Define.
949 (LD_WHOLE_ARCHIVE_OPTION): Define.
950 (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
951 (ASAN_REJECT_SPEC): Provide default.
952 (LIBASAN_EARLY_SPEC): Define.
953 (LIBTSAN_EARLY_SPEC): Define.
954 (LIBLSAN_EARLY_SPEC): Define.
955 * config/i386/sol2.h (CC1_SPEC): Redefine.
956 (ASAN_REJECT_SPEC): Define.
958 * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
959 (TARGET_ASAN_SHADOW_OFFSET): Define.
960 (sparc_asan_shadow_offset): New function.
961 * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
962 (ASAN_REJECT_SPEC): Define.
964 2018-11-06 Jan Hubicka <jh@suse.cz>
966 * tree.c (fld_type_variant): Copy canonical type.
967 (fld_incomplete_type_of): Check that canonical types looks sane;
969 (verify_type): Accept when incomplete type has complete canonical type.
971 2018-11-06 Jan Hubicka <jh@suse.cz>
973 * tree.c (free_lang_data): Reset overwite_assembler_name,
974 print_xnode, print_decl, print_type and print_identifier of
977 2018-11-06 Richard Biener <rguenther@suse.de>
979 PR tree-optimization/87889
980 * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
981 Do nothing if old and new arg are the same
983 2018-11-06 Andreas Krebbel <krebbel@linux.ibm.com>
986 * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
987 attributes for operands 3 and 4.
989 2018-11-06 Richard Biener <rguenther@suse.de>
992 * simplify-rtx.c (simplify_binary_operation_1): Add pattern
993 matching bitfield insertion.
995 2018-11-06 Alexandre Oliva <aoliva@redhat.com>
997 * auto-inc-dec.c: Include valtrack.h. Improve comments.
998 (reg_next_debug_use): New.
999 (attempt_change): Propagate adjusted expression into affected
1001 (merge_in_block): Track uses in debug insns.
1002 (pass_inc_dec::execute): Allocate and release
1005 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
1007 * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
1008 (_mm512_fixupimm_round_pd): Update parameters and builtin.
1009 (_mm512_maskz_fixupimm_round_pd): Ditto.
1010 (_mm512_fixupimm_round_ps): Ditto.
1011 (_mm512_maskz_fixupimm_round_ps): Ditto.
1012 (_mm_fixupimm_round_sd): Ditto.
1013 (_mm_maskz_fixupimm_round_sd): Ditto.
1014 (_mm_fixupimm_round_ss): Ditto.
1015 (_mm_maskz_fixupimm_round_ss): Ditto.
1016 (_mm512_fixupimm_pd): Ditto.
1017 (_mm512_maskz_fixupimm_pd): Ditto.
1018 (_mm512_fixupimm_ps): Ditto.
1019 (_mm512_maskz_fixupimm_ps): Ditto.
1020 (_mm_fixupimm_sd): Ditto.
1021 (_mm_maskz_fixupimm_sd): Ditto.
1022 (_mm_fixupimm_ss): Ditto.
1023 (_mm_maskz_fixupimm_ss): Ditto.
1024 (_mm512_mask_fixupimm_round_pd): Update builtin.
1025 (_mm512_mask_fixupimm_round_ps): Ditto.
1026 (_mm_mask_fixupimm_round_sd): Ditto.
1027 (_mm_mask_fixupimm_round_ss): Ditto.
1028 (_mm512_mask_fixupimm_pd): Ditto.
1029 (_mm512_mask_fixupimm_ps): Ditto.
1030 (_mm_mask_fixupimm_sd): Ditto.
1031 (_mm_mask_fixupimm_ss): Ditto.
1032 * config/i386/avx512vlintrin.h:
1033 (_mm256_fixupimm_pd): Update parameters and builtin.
1034 (_mm256_maskz_fixupimm_pd): Ditto.
1035 (_mm256_fixupimm_ps): Ditto.
1036 (_mm256_maskz_fixupimm_ps): Ditto.
1037 (_mm_fixupimm_pd): Ditto.
1038 (_mm_maskz_fixupimm_pd): Ditto.
1039 (_mm_fixupimm_ps): Ditto.
1040 (_mm_maskz_fixupimm_ps): Ditto.
1041 (_mm256_mask_fixupimm_pd): Update builtin.
1042 (_mm256_mask_fixupimm_ps): Ditto.
1043 (_mm_mask_fixupimm_pd): Ditto.
1044 (_mm_mask_fixupimm_ps): Ditto.
1045 * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
1046 * config/i386/i386-builtin.def: Update builtin definitions.
1047 * config/i386/i386.c: Handle new builtin types and remove useless ones.
1048 * config/i386/sse.md: Update VFIXUPIMM* patterns.
1049 (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1050 (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1051 (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
1052 (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1053 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1054 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
1055 * config/i386/subst.md:
1056 (round_saeonly_sd_mask_operand4): Add new subst_attr.
1057 (round_saeonly_sd_mask_op4): Ditto.
1058 (round_saeonly_expand_operand5): Ditto.
1059 (round_saeonly_expand): Update.
1061 2018-11-05 Max Filippov <jcmvbkbc@gmail.com>
1063 * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
1065 2018-11-05 Segher Boessenkool <segher@kernel.crashing.org>
1067 PR rtl-optimization/87871
1068 * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
1070 2018-11-05 Paul Koning <ni1d@arrl.net>
1072 * doc/sourcebuild.texi (target attributes): Document new "inf"
1073 effective target keyword.
1075 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
1077 * config/s390/s390.c (s390_register_move_cost): Increase costs for
1078 moves involving the CC reg.
1080 2018-11-05 Richard Biener <rguenther@suse.de>
1082 PR tree-optimization/87873
1083 * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
1085 * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
1086 * tree-vect-loop.c (vect_transform_loop): When splitting the
1087 loop exit also create forwarder PHIs for constants.
1088 * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
1089 Handle constant to_arg, add extra checking we match up the correct
1092 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
1094 * config/s390/s390.md: QImode and HImode for load on condition.
1096 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
1098 * config/s390/predicates.md: Fix typo.
1099 * config/s390/s390.md: Allow immediates for load on condition.
1101 2018-11-05 Martin Liska <mliska@suse.cz>
1103 * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
1104 * bitmap.h (struct bitmap_usage): Likewise.
1105 * ggc-common.c (SCALE): Remove.
1107 (struct ggc_usage): Use SIZE_AMOUNT. And update
1109 * ggc-page.c (SCALE): Remove.
1110 (STAT_LABEL): Remove.
1111 (ggc_print_statistics): Use SIZE_AMOUNT.
1112 * gimple.h (SCALE): Remove.
1114 * input.c (ONE_K): Remove.
1117 (STAT_LABEL): Likewise.
1118 (FORMAT_AMOUNT): Likewise.
1119 (dump_line_table_statistics): Use SIZE_AMOUNT.
1120 * mem-stats.h (struct mem_usage): Likewise.
1121 * rtl.c (dump_rtx_statistics): Likewise.
1122 (rtx_alloc_counts): Change type to size_t.
1123 (rtx_alloc_sizes): Likewise.
1124 (rtx_count_cmp): New.
1125 (dump_rtx_statistics): Sort first based on counts.
1126 * tree.c (tree_nodes_cmp): New.
1127 (tree_codes_cmp): New.
1128 (dump_tree_statistics): Sort first based on counts.
1129 * system.h (ONE_K): New.
1131 (SIZE_SCALE): Likewise.
1132 (SIZE_LABEL): Likewise.
1133 (SIZE_AMOUNT): Likewise.
1134 * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
1135 * tree-dfa.c (dump_dfa_stats): Likewise.
1136 * tree-phinodes.c (phinodes_print_statistics): Likewise.
1137 * tree-ssanames.c (ssanames_print_statistics): Likewise.
1138 * tree.c (dump_tree_statistics): Likewise.
1139 * vec.c (struct vec_usage): Likewise.
1140 * trans-mem.c (tm_mangle): Enlarge buffer in order to not
1141 trigger a -Werror=format-overflow with
1142 --enable-gather-detailed-stats.
1144 2018-11-05 Martin Liska <mliska@suse.cz>
1146 * mem-stats.h (mem_alloc_description::release_instance_overhead):
1148 * vec.c (struct vec_usage): Register m_element_size.
1149 (vec_prefix::register_overhead): New arguments: elements and
1151 (vec_prefix::release_overhead): Subtract elements.
1152 * vec.h (struct vec_prefix): Change signature.
1153 (va_heap::reserve): Pass proper arguments.
1154 (va_heap::release): Likewise.
1156 2018-11-05 Martin Liska <mliska@suse.cz>
1158 * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
1162 2018-11-05 Richard Biener <rguenther@suse.de>
1164 * tree-scalar-evolution.h (final_value_replacement_loop): Update
1166 * tree-scalar-evolution.c (final_value_replacement_loop): Return
1167 whether anything was done.
1168 (scev_const_prop): Remove constant propagation part, fold
1170 * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
1171 (pass_data_scev_cprop): TODO_cleanup_cfg is now done
1174 2018-11-05 Jakub Jelinek <jakub@redhat.com>
1176 PR tree-optimization/87859
1177 * gimple-ssa-store-merging.c (struct merged_store_group): Add
1178 only_constants and first_nonmergeable_order members.
1179 (merged_store_group::merged_store_group): Initialize them.
1180 (merged_store_group::do_merge): Clear only_constants member if
1181 adding something other than INTEGER_CST store.
1182 (imm_store_chain_info::coalesce_immediate_stores): Don't merge
1183 stores with order >= first_nonmergeable_order. Use
1184 merged_store->only_constants instead of always recomputing it.
1185 Set merged_store->first_nonmergeable_order if we've skipped any
1186 stores. Attempt to merge overlapping INTEGER_CST stores that
1187 we would otherwise skip.
1190 * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
1192 2018-11-05 Xuepeng Guo <xuepeng.guo@intel.com>
1195 * config/i386/emmintrin.h (__v16qs): New to cope with option
1197 (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
1198 (_mm_cmplt_epi8): Likewise.
1199 (_mm_cmpgt_epi8): Likewise.
1201 2018-11-05 Richard Biener <rguenther@suse.de>
1203 PR rtl-optimization/87852
1204 * fwprop.c (use_killed_between): Only consider single-defs of the
1205 use whose definition statement dominates the use.
1207 2018-11-05 Martin Liska <mliska@suse.cz>
1210 * doc/invoke.texi: Remove options that are
1211 not disabled with -Os.
1213 2018-11-05 Martin Liska <mliska@suse.cz>
1216 * doc/extend.texi: Update constrain about the last argument
1217 of __builtin_expect_with_probability.
1219 2018-11-05 Martin Liska <mliska@suse.cz>
1222 * predict.c (expr_expected_value_1): Verify
1223 that last argument is a real constants and emit
1226 2018-11-05 Martin Liska <mliska@suse.cz>
1228 PR gcov-profile/77698
1229 * ipa-profile.c (ipa_profile): Adjust hotness threshold
1232 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
1234 PR tree-optimization/86572
1235 * builtins.c (c_strlen): Handle negative offsets in a safe way.
1237 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
1239 PR tree-optimization/87672
1240 * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
1241 * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
1243 2018-11-04 Uros Bizjak <ubizjak@gmail.com>
1246 * cfgexpand.c (pass_expand::execute): Move the call to
1247 finish_eh_generation in front of the call to expand_stack_alignment.
1249 2018-11-04 Venkataramanan Kumar <venkataramanan.kumar@amd.com>
1251 * common/config/i386/i386-common.c (processor_alias_table): Add
1253 * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
1254 (case ${target}): Add znver2.
1255 * config/i386/driver-i386.c: (host_detect_local_cpu): Let
1256 -march=native recognize znver2 processors.
1257 * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
1258 * config/i386/i386.c (m_znver2): New definition.
1259 (m_ZNVER): New definition.
1260 (m_AMD_MULTIPLE): Includes m_znver2.
1261 (processor_cost_table): Add znver2 entry.
1262 (processor_target_table): Add znver2 entry.
1263 (get_builtin_code_for_version): Set priority for
1265 (processor_model): Add M_AMDFAM17H_ZNVER2.
1266 (arch_names_table): Ditto.
1267 (ix86_reassociation_width): Include znver2.
1268 * config/i386/i386.h (TARGET_znver2): New definition.
1269 (struct ix86_size_cost): Add TARGET_ZNVER2.
1270 (enum processor_type): Add PROCESSOR_ZNVER2.
1271 * config/i386/i386.md (define_attr "cpu"): Add znver2.
1272 * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
1273 * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
1274 (ix86_adjust_cost): Add znver2.
1275 * config/i386/x86-tune.def: Replace m_ZNVER1 by m_ZNVER.
1276 * gcc/doc/extend.texi: Add details about znver2.
1277 * gcc/doc/invoke.texi: Add details about znver2.
1279 2018-11-03 Sandra Loosemore <sandra@codesourcery.com>
1283 * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
1286 2018-11-02 Aaron Sawdey <acsawdey@linux.ibm.com>
1288 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
1289 attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
1291 2018-11-02 Richard Earnshaw <rearnsha@arm.com>
1293 * config/aarch64/aarch64.c ((aarch64_override_options): Disable
1294 shrink-wrapping when -mtrack-speculation.
1296 2018-11-02 Richard Biener <rguenther@suse.de>
1298 * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
1299 (pop_cost_one_pair): Do not free pair.
1300 (pop_best_coalesce): Likewise.
1301 (create_coalesce_list): Initialize obstack.
1302 (delete_coalesce_list): Free obstack.
1303 (find_coalesce_pair): Obstack-allocate coalesce pairs.
1304 (add_cost_one_coalesce): Likewise.
1305 (struct live_track): Remove bitmap pointer indirections.
1306 (new_live_track): Adjust.
1307 (delete_live_track): Likewise.
1308 (live_track_remove_partition): Likewise.
1309 (live_track_add_partition): Likewise.
1310 (live_track_live_p): Likewise.
1311 (live_track_process_def): Likewise.
1312 (live_track_clear_base_vars): Likewise.
1314 2018-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1316 * configure.ac (gcc_cv_as_sparc_register_op): Remove.
1317 * configure: Regenerate.
1318 * config.in: Regenerate.
1319 * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
1320 HAVE_AS_REGISTER_PSEUDO_OP guard.
1321 * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
1322 !HAVE_AS_REGISTER_PSEUDO_OP support.
1323 (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
1326 2018-11-02 Richard Biener <rguenther@suse.de>
1328 * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
1331 2018-11-02 Richard Biener <rguenther@suse.de>
1333 PR tree-optimization/87776
1334 * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
1335 executable when iterating but running into rpo-vn-max-loop-depth
1336 and not eliding the iteration.
1338 2018-11-30 Jan Hubicka <jh@suse.cz>
1340 * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
1343 2018-11-01 Aaron Sawdey <acsawdey@linux.ibm.com>
1345 * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
1347 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
1348 Combine with rs6000_address_for_fpconvert.
1349 (rs6000_address_for_fpconvert): Combine with
1350 rs6000_force_indexed_or_indirect_mem.
1351 (rs6000_expand_vector_init): Change function call from
1352 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
1353 * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
1354 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
1355 (floatsi<mode>2_lfiwax_mem): Ditto.
1356 (floatunssi<mode>2_lfiwzx): Ditto.
1357 (floatunssi<mode>2_lfiwzx_mem): Ditto.
1358 (float<QHI:mode><FP_ISA3:mode>2): Ditto.
1359 (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
1360 (fix_trunc<mode>si2_stfiwx): Ditto.
1361 (fixuns_trunc<mode>si2_stfiwx): Ditto.
1362 (float_<mode>si2_hw): Ditto.
1363 (floatuns_<mode>si2_hw): Ditto.
1364 * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
1365 (vsx_splat_<mode>): Ditto.
1367 2018-11-01 Joseph Myers <joseph@codesourcery.com>
1369 * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
1370 at end of assembler input text.
1371 * configure: Regenerate.
1373 2018-11-01 Jakub Jelinek <jakub@redhat.com>
1375 PR tree-optimization/87826
1376 * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
1377 negative or larger or equal to type's precision.
1379 2018-10-31 Alexandre Oliva <aoliva@redhat.com>
1381 * opts.c (default_options_table): Do not enable
1382 OPT_fdelayed_branch at -Og.
1383 * doc/invoke.texi (-fdelayed-branch): Document it.
1385 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
1387 * optabs-libfuncs.c (build_libfunc_function_visibility):
1388 New, split out from...
1389 (build_libfunc_function): ... here.
1390 (init_one_libfunc_visibility): New, split out from ...
1391 (init_one_libfunc): ... here.
1393 * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
1394 scratch register need not be early-clobber. Document the reason
1395 why we cannot use ST<OP>.
1397 2018-10-31 Joseph Myers <joseph@codesourcery.com>
1400 * configure.ac: Remove AC_PREREQ. Use AC_LANG_SOURCE. Use single
1401 line for second argument of AC_DEFINE_UNQUOTED.
1402 * doc/install.texi (Tools/packages necessary for modifying GCC):
1403 Update to autoconf 2.69 and automake 1.15.1.
1404 * aclocal.m4, config.in, configure: Regenerate.
1406 2018-10-31 Pat Haugen <pthaugen@us.ibm.com>
1408 * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
1411 2018-10-31 Martin Liska <mliska@suse.cz>
1414 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
1415 Add new argument invalid_extension.
1416 (aarch64_get_all_extension_candidates): New function.
1417 (aarch64_rewrite_selected_cpu): Add NULL to function call.
1418 * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
1420 (aarch64_get_all_extension_candidates): New function.
1421 * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
1422 argument invalid_extension.
1423 (aarch64_parse_cpu): Likewise.
1424 (aarch64_print_hint_for_extensions): New function.
1425 (aarch64_validate_mcpu): Provide hint about invalid extension.
1426 (aarch64_validate_march): Likewise.
1427 (aarch64_handle_attr_arch): Pass new argument.
1428 (aarch64_handle_attr_cpu): Provide hint about invalid extension.
1429 (aarch64_handle_attr_isa_flags): Likewise.
1431 2018-10-31 Richard Biener <rguenther@suse.de>
1435 * tree-outof-ssa.c (insert_backedge_copies): Restrict
1436 copy generation to useful cases. Place the copy before
1437 the definition of the backedge value when possible.
1439 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
1441 * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
1442 * config/arc/arc.c (arc_active_insn): New function.
1443 (check_store_cacheline_hazard): Likewise.
1444 (workaround_arc_anomaly): Use check_store_cacheline_hazard.
1445 (arc_override_options): Disable delay slot scheduler for older
1447 (arc_store_addr_hazard_p): New implementation, old one renamed to
1449 (arc_store_addr_hazard_internal_p): Renamed.
1450 (arc_reorg): Don't combine into brcc instructions which are part
1451 of hardware hazard solution.
1452 * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
1453 (tune_arc700): Likewise.
1454 * config/arc/arc.opt (arc7xx): New tune value.
1455 * config/arc/arc700.md: Improve A7 scheduler.
1457 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
1459 * config/arc/arc.c (arc_override_options): Remove
1460 TARGET_COMPACT_CASESI.
1461 * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
1462 (CASE_VECTOR_MODE): Likewise.
1463 (CASE_VECTOR_PC_RELATIVE): Likewise.
1464 (CASE_VECTOR_SHORTEN_MODE): Likewise.
1465 (CASE_VECTOR_SHORTEN_MODE1): Delete.
1466 (ADDR_VEC_ALIGN): Update.
1467 (ASM_OUTPUT_CASE_LABEL): Undefine.
1468 (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
1469 (TARGET_BI_BIH): Define.
1470 (DEFAULT_BRANCH_INDEX): Likewise.
1471 * config/arc/arc.md (casesi): Rework to accept BI/BIH
1472 instructions, remove compact_casesi use case.
1473 (casesi_compact_jump): Remove.
1474 (casesi_dispatch): New pattern.
1475 * config/arc/arc.opt: Add mbranch-index option. Deprecate
1476 compact_casesi option.
1477 * doc/invoke.texi: Document mbranch-index option.
1479 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
1481 * config/arc/arc.c (arc_get_tp): Remove function.
1482 (arc_emit_call_tls_get_addr): Likewise.
1483 (arc_call_tls_get_addr): New function.
1484 (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
1485 * config/arc/arc.md (tls_load_tp_soft): Remove.
1486 (tls_gd_get_addr): Likewise.
1488 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
1490 * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
1491 (mulsi3_600_lib): Remove pattern.
1492 (umulsi3_highpart_600_lib_le): Likewise.
1493 (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
1494 (umulsidi3): Remove call to umulsidi3_600_lib.
1495 (umulsidi3_600_lib): Remove pattern.
1496 (peephole2): Remove peephole using the above deprecated patterns.
1498 2018-10-31 Thomas Preud'homme <thomas.preudhomme@linaro.org>
1501 * config/arm/arm.c (arm_option_check_internal): Disable the combined
1502 use of -mslow-flash-data and -mword-relocations.
1503 (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
1504 * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
1506 * doc/invoke.texi (-mword-relocations): Mention conflict with
1508 (-mslow-flash-data): Reciprocally.
1510 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
1512 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
1513 16-byte modes held in GP registers to use an even regno.
1515 * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
1516 (aarch64_atomic_ldop_supported_p): Remove.
1517 (aarch64_gen_atomic_ldop): Remove.
1518 * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
1519 Fully expand LSE operations here.
1520 (atomic_fetch_<atomic_optab><ALLI>): Likewise.
1521 (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
1522 (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
1523 and use ATOMIC_LDOP instead; use register_operand for the input;
1524 drop the split and emit insns directly.
1525 (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
1526 (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
1527 (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
1529 * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
1530 (aarch64_gen_atomic_ldop): Don't call it.
1531 * config/aarch64/atomics.md (atomic_exchange<ALLI>):
1532 Use aarch64_reg_or_zero.
1533 (aarch64_atomic_exchange<ALLI>): Likewise.
1534 (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
1535 operand 0; use aarch64_reg_or_zero for input; merge ...
1536 (@aarch64_atomic_swp<ALLI>): ... this and remove.
1538 * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
1539 (aarch64_split_compare_and_swap): Use it.
1540 (aarch64_expand_compare_and_swap): Likewise. Remove convert_modes;
1541 test oldval against the proper predicate.
1542 * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
1543 Use nonmemory_operand for expected.
1544 (cas_short_expected_pred): New.
1545 (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
1546 (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
1547 * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
1548 (aarch64_plushi_operand): New.
1550 * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
1551 Force oldval into the rval register for TARGET_LSE; emit the compare
1552 during initial expansion so that it may be deleted if unused.
1553 (aarch64_gen_atomic_cas): Remove.
1554 * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
1555 Change =&r to +r for operand 0; use match_dup for operand 2;
1556 remove is_weak and mod_f operands as unused. Drop the split
1558 (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
1559 (@aarch64_compare_and_swap<GPI>_lse): Similarly.
1560 (@aarch64_atomic_cas<GPI>): Similarly.
1562 2018-10-31 Richard Biener <rguenther@suse.de>
1564 * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
1567 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1569 * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
1570 * config/aarch64/aarch64.md: Include saphira.md
1571 * config/aarch64/saphira.md: New file for pipeline description.
1573 2018-10-30 Martin Sebor <msebor@redhat.com>
1576 * gimple-ssa-sprintf.c (format_directive): Use %G to include
1578 (sprintf_dom_walker::compute_format_length):
1579 Avoid setting POSUNDER4K here.
1580 (get_destination_size): Handle null argument values.
1581 (get_user_idx_format): New function.
1582 (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
1583 functions, including user-defined with attribute format printf.
1584 Use %G to include inlining context.
1585 Set POSUNDER4K here.
1587 2018-10-30 Jan Hubicka <jh@suse.cz>
1589 * params.def (lto-partitions): Bump from 32 to 128.
1591 2018-10-30 Jan Hubicka <jh@suse.cz>
1594 (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
1596 (free_lang_data_in_type): Forward declare.
1597 (fld_type_variant_equal_p): New function.
1598 (fld_type_variant): New function
1599 (fld_incomplete_types): New hash.
1600 (fld_incomplete_type_of): New function
1601 (fld_simplfied-type): New function.
1602 (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
1603 (free_lang_data): Allocate and free fld_incomplete_type; update call
1604 of free_lang_data_in_decl.
1606 2018-10-30 Eric Botcazou <ebotcazou@adacore.com>
1608 * gcov.c (output_lines): Remove duplicate line.
1610 2018-10-30 Aaron Sawdey <acsawdey@linux.ibm.com>
1612 * config/rs6000/rs6000.md (bswapdi2): Force address into register
1613 if not in indexed or indirect form.
1614 (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
1615 (bswapdi2_store): Ditto.
1616 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
1618 * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
1619 Prototype for helper function.
1621 2018-10-30 Martin Sebor <msebor@redhat.com>
1623 * doc/extend.texi (optimize): Clarify/expand attribute documentation.
1624 (target, pragma GCC optimize, pragma GCC target): Ditto.
1626 2018-10-30 Jonathan Wakely <jwakely@redhat.com>
1628 * doc/extend.texi: Fix prototype and description of
1629 __builtin_expect_with_probability.
1631 2018-10-30 Michael Ploujnikov <michael.ploujnikov@oracle.com>
1633 * cgraph.h (clone_function_name_1): Replaced by new
1634 clone_function_name_numbered that takes name as string; for
1635 privatize_symbol_name_1 use only.
1636 (clone_function_name): Renamed to
1637 clone_function_name_numbered to be explicit about numbering.
1638 (clone_function_name): New two-argument function that does
1639 not number its output.
1640 (clone_function_name): New three-argument function that
1641 takes a number to append to its output.
1642 * cgraphclones.c (duplicate_thunk_for_node):
1643 (clone_function_name_1): Renamed.
1644 (clone_function_name_numbered): Two new functions.
1645 (clone_function_name): Improved documentation.
1646 (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
1647 * config/rs6000/rs6000.c (make_resolver_func): Ditto.
1648 * final.c (final_scan_insn_1): Use the new clone_function_name
1650 * multiple_target.c (create_dispatcher_calls): Ditto.
1651 (create_target_clone): Ditto.
1652 * omp-expand.c (grid_expand_target_grid_body): Ditto.
1653 * omp-low.c (create_omp_child_function_name): Ditto.
1654 * omp-simd-clone.c (simd_clone_create): Ditto.
1655 * symtab.c (simd_symtab_node::noninterposable_alias): Use the
1656 new clone_function_name without numbering.
1658 2018-10-30 Richard Earnshaw <rearnsha@arm.com>
1660 * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
1661 Assert that the allocation size is not zero.
1663 2018-10-30 Richard Biener <rguenther@suse.de>
1665 PR tree-optimization/87800
1666 * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
1667 non-induction or reduction PHIs.
1669 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
1671 * config/aarch64/falkor-tag-collision-avoidance.c
1672 (execute_tag_collision_avoidance): Call df_note_add_problem.
1674 2018-10-30 Martin Liska <mliska@suse.cz>
1676 * doc/extend.texi: Fix typo in documentation
1677 of __builtin_expect_with_probability.
1679 2018-10-29 David Malcolm <dmalcolm@redhat.com>
1682 * input.c (get_substring_ranges_for_loc): Detect if
1683 linemap_resolve_location gives us a NULL map, and reject
1686 2018-10-29 Iain Buclaw <ibuclaw@gdcproject.org>
1688 * config.gcc (xstormy16-*-elf): Set tm_d_file.
1690 2018-10-29 David Malcolm <dmalcolm@redhat.com>
1691 Martin Sebor <msebor@redhat.com>
1692 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1694 * doc/ux.texi (Quoting): New subsection, adapted from material at
1695 https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
1696 MartinSebor and ManuelLopezIbanez.
1697 (Fix-it hints): Note that fix-it hints shouldn't be marked for
1700 2018-10-29 Kugan Vivekanandarajah <kuganv@linaro.org>
1703 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
1706 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
1708 * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
1710 2018-10-29 Paul Koning <ni1d@arrl.net>
1712 * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
1714 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
1716 * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
1717 _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
1718 _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
1719 _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
1720 _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
1721 _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
1722 _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
1723 _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
1724 Change 'vector' to '__vector'.
1725 * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
1726 _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
1727 _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
1728 _mm_avg_pu16): Likewise. And, whitespace corrections.
1730 2018-10-29 Richard Biener <rguenther@suse.de>
1732 PR tree-optimization/87785
1733 * tree-vect-slp.c (vect_gather_slp_loads): Only gather
1736 2018-10-29 Olivier Hainque <hainque@adacore.com>
1738 * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
1740 2018-10-29 Olivier Hainque <hainque@adacore.com>
1742 * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
1743 ports configuration macro, defaults to "gnu".
1744 (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
1745 TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
1747 2018-10-29 Olivier Hainque <hainque@adacore.com>
1749 * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
1751 2018-10-29 Richard Biener <rguenther@suse.de>
1753 PR tree-optimization/87790
1754 * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
1755 (vect_make_slp_decision): Adjust.
1756 (vect_slp_analyze_bb_1): Likewise.
1757 (vect_detect_hybrid_slp_stmts): Properly union SLP type over
1760 2018-10-29 Richard Biener <rguenther@suse.de>
1762 PR tree-optimization/87785
1763 * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
1765 (vect_build_slp_tree): Likewise.
1766 (vect_gather_slp_loads): New function.
1767 (vect_analyze_slp_instance): Gather loads separately from the
1770 2018-10-29 Martin Liska <mliska@suse.cz>
1772 * Makefile.in: Make dependency to json.o.
1773 * doc/gcov.texi: Document new JSON format, remove
1774 old intermediate format documentation.
1775 * gcov.c (struct function_info): Come up with m_name and
1777 (function_info::function_info): Initialize it.
1778 (function_info::~function_info): Release it.
1779 (main): Rename flag_intermediate_format to flag_json_format.
1780 (print_usage): Describe --json-format.
1781 (process_args): Set flag_json_format.
1782 (output_intermediate_line): Remove.
1783 (output_intermediate_json_line): Likewise.
1784 (get_gcov_intermediate_filename): Return new extension
1786 (output_intermediate_file): Implement JSON emission.
1787 (output_json_intermediate_file): Implement JSON emission.
1788 (generate_results): Use ::get_name for function name.
1789 Handle JSON output file.
1790 (read_graph_file): Use ::get_name instead of cplus_demangle.
1791 (read_count_file): Likewise.
1792 (solve_flow_graph): Likewise.
1793 (add_line_counts): Likewise.
1794 (accumulate_line_counts): Use new flag_json_format.
1795 (output_function_details): Use ::get_name instead of cplus_demangle.
1796 (output_lines): Likewise.
1797 * json.cc (test_writing_literals): Add new tests.
1798 * json.h (class literal): Add new boolean constructor.
1800 2018-10-29 Segher Boessenkool <segher@kernel.crashing.org>
1802 PR rtl-optimization/87701
1803 PR rtl-optimization/87780
1804 * combine.c (make_more_copies): Rewrite.
1806 2018-10-28 Kugan Vivekanandarajah <kuganv@linaro.org>
1808 * doc/generic.texi (ABSU_EXPR): Document.
1809 * match.pd (absu(x)*absu(x) -> x*x): Handle.
1810 (absu(absu(X)) -> absu(X)): Likewise.
1811 (absu(-X) -> absu(X)): Likewise.
1812 (absu(X) where X is nonnegative -> X): Likewise.
1814 2018-10-28 Iain Buclaw <ibuclaw@gdcproject.org>
1816 * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
1817 (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
1818 (tm_d.h, cs-tm_d.h, default-d.o): New rules.
1819 (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
1820 (s-tm-texi): Also check timestamp on d-target.def.
1821 (generated_files): Add TM_D_H and d-target-hooks-def.h.
1822 (build/genhooks.o): Also depend on D_TARGET_DEF.
1823 * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
1825 * config/aarch64/aarch64-d.c: New file.
1826 * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
1828 * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
1830 * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
1831 * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
1832 * config/arm/arm-d.c: New file.
1833 * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
1834 * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
1835 * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
1836 * config/arm/t-arm (arm-d.o): New rule.
1837 * config/default-d.c: New file.
1838 * config/glibc-d.c: New file.
1839 * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1840 * config/i386/i386-d.c: New file.
1841 * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
1842 * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
1843 * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
1844 (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
1845 * config/i386/t-i386 (i386-d.o): New rule.
1846 * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1847 * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1848 * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
1849 * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1850 * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
1851 * config/mips/mips-d.c: New file.
1852 * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
1853 * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
1854 * config/mips/t-mips (mips-d.o): New rule.
1855 * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1856 * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1857 * config/powerpcspe/powerpcspe-d.c: New file.
1858 * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
1860 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
1861 Support GNU D by using 0 as the language type.
1862 * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
1863 * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
1864 * config/riscv/riscv-d.c: New file.
1865 * config/riscv/riscv-protos.h (riscv_d_target_versions): New
1867 * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
1868 * config/riscv/t-riscv (riscv-d.o): New rule.
1869 * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1870 * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1871 * config/rs6000/rs6000-d.c: New file.
1872 * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
1874 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
1875 Support GNU D by using 0 as the language type.
1876 * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
1877 * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
1878 * config/s390/s390-d.c: New file.
1879 * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
1880 * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
1881 * config/s390/t-s390 (s390-d.o): New rule.
1882 * config/sparc/sparc-d.c: New file.
1883 * config/sparc/sparc-protos.h (sparc_d_target_versions): New
1885 * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
1886 * config/sparc/t-sparc (sparc-d.o): New rule.
1887 * config/t-glibc (glibc-d.o): New rule.
1888 * configure: Regenerated.
1889 * configure.ac (tm_d_file): New variable.
1890 (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
1891 * doc/contrib.texi (Contributors): Add self for the D frontend.
1892 * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
1893 * doc/install.texi (Configuration): Mention libphobos as an option for
1894 --enable-shared. Mention d as an option for --enable-languages.
1895 (Testing): Mention check-d as a target.
1896 * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
1897 name suffixes. Mention d as a -x option.
1898 * doc/sourcebuild.texi (Top Level): Mention libphobos.
1899 * doc/standards.texi (Standards): Add section on D language.
1900 * doc/tm.texi: Regenerated.
1901 * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
1902 TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
1903 * dwarf2out.c (is_dlang): New function.
1904 (gen_compile_unit_die): Use DW_LANG_D for D.
1905 (declare_in_namespace): Return module die for D, instead of adding
1906 extra declarations into the namespace.
1907 (gen_namespace_die): Generate DW_TAG_module for D.
1908 (gen_decl_die): Handle CONST_DECLSs for D.
1909 (dwarf2out_decl): Likewise.
1910 (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
1911 (prune_unused_types_walk): Handle DW_tag_interface_type same as other
1912 kinds of aggregates.
1913 * gcc.c (default_compilers): Add entries for .d, .dd and .di.
1914 * genhooks.c: Include d/d-target.def.
1916 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
1919 * config/rs6000/darwin.h (STACK_BOUNDARY): New.
1920 (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
1921 (STACK_DYNAMIC_OFFSET): Likewise.
1923 2018-10-27 Sandra Loosemore <sandra@codesourcery.com>
1926 * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
1929 2018-10-26 Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com>
1931 * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
1932 return type and other typos.
1934 2018-10-26 Aaron Sawdey <acsawdey@linux.ibm.com>
1936 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
1937 a shorter sequence with fewer branches.
1938 (emit_final_str_compare_gpr): Ditto.
1940 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
1942 * config/rs6000/tmmintrin.h: New file.
1943 * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
1945 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
1947 * config/rs6000/mmintrin.h: Enable 32bit compilation.
1948 * config/rs6000/xmmintrin.h: Likewise.
1950 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
1952 * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
1954 2018-10-26 Richard Biener <rguenther@suse.de>
1956 * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
1958 (vect_mark_slp_stmts_relevant): Likewise.
1959 (vect_detect_hybrid_slp_stmts): Likewise.
1960 (vect_bb_slp_scalar_cost): Likewise.
1961 (vect_remove_slp_scalar_calls): Likewise.
1963 2018-10-26 Jan Hubicka <jh@suse.cz>
1965 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
1966 (warn_types_mismatch): Fix walk of DECL_NAME.
1967 (odr_types_equivalent_p): Fix overactive assert.
1969 2018-10-26 Richard Biener <rguenther@suse.de>
1971 PR tree-optimization/87105
1972 * tree-vectorizer.h (_slp_tree::refcnt): New member.
1973 * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
1975 (vect_create_new_slp_node): Initialize refcnt to one.
1977 (scalar_stmts_set_t, bst_fail): Remove.
1978 (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
1979 (vect_build_slp_tree): Add bst_map argument and lookup
1980 already created SLP nodes.
1981 (vect_print_slp_tree): Handle a SLP graph, print SLP node
1983 (vect_slp_rearrange_stmts): Handle a SLP graph.
1984 (vect_analyze_slp_instance): Adjust and free SLP nodes from
1985 the CSE map. Fix indenting.
1986 (vect_schedule_slp_instance): Add short-cut.
1988 2018-10-26 Martin Liska <mliska@suse.cz>
1991 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
1992 addr_expr and not with pointers.
1994 2018-10-26 Jan Hubicka <jh@suse.cz>
1996 * tree.c (free_lang_data_in_type): Only check main variants.
1997 * ipa-devirt.c (warn_odr): Make static.
1998 (types_same_for_odr): Drop strict variant.
1999 (types_odr_comparable): Likewise.
2000 (odr_or_derived_type_p): Look for main variants.
2001 (odr_name_hasher::equal): Cleanup comment.
2002 (odr_subtypes_equivalent): Add warn and warned arguments; check main
2004 (type_variants_equivalent_p): break out from ...
2005 (odr_types_equivalent): ... here; go for main variants where needed.
2006 (warn_odr): ... here; turn static.
2007 (warn_types_mismatch): Compare mangled names of main variants.
2008 * ipa-utils.h (types_odr_comparable): Drop strict parameter.
2009 (type_with_linkage_p): Sanity check that we look at main variant.
2010 * lto.c (lto_read_decls): Only consider main variant to be ODR type.
2011 * tree.h (types_same_for_odr): Drop strict argument.
2013 2018-10-26 Richard Biener <rguenther@suse.de>
2015 PR tree-optimization/87746
2016 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
2017 Simplify and fix WRT strided store groups with size not
2018 equal to step in element count.
2019 (vect_analyze_group_access_1): Dump the whole group.
2021 2018-10-25 Carl Love <cel@us.ibm.com>
2023 * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
2024 P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
2025 P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name. Add quad
2026 precicion entry for each overloaded builtin.
2027 * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
2028 VSCEDPUO): Rename overloaded name.
2029 (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
2030 VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
2031 * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
2032 define_expand for xscmpexqp instruction.
2033 (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
2035 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
2036 Jinsong Ji <jji@us.ibm.com>
2038 * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
2039 function with vec_sl.
2040 (_mm_slli_epi32): Likewise.
2041 (_mm_slli_epi64): Likewise.
2042 (_mm_srai_epi16): Replace deprecated function with vec_sra.
2043 (_mm_srai_epi32): Likewise.
2044 (_mm_srli_epi16): Replace deprecated function with vec_sr.
2045 (_mm_srli_epi32): Likewise.
2046 (_mm_srli_epi64): Likewise.
2047 (_mm_sll_epi16): Replace deprecated function with vec_sl.
2048 (_mm_sll_epi32): Likewise.
2049 (_mm_sll_epi64): Likewise.
2050 (_mm_sra_epi16): Replace deprecated function with vec_sra.
2051 (_mm_sra_epi32): Likewise.
2052 (_mm_srl_epi16): Replace deprecated function with vec_sr.
2053 (_mm_srl_epi32): Likewise.
2054 (_mm_srl_epi64): Likewise.
2056 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
2057 Jinsong Ji <jji@us.ibm.com>
2059 * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
2060 comparison operators with vec_cmp* for compatibility due to
2061 unfortunate history; clean up formatting and use types more
2063 (_mm_sll_epi32): Likewise.
2064 (_mm_sll_epi64): Likewise.
2065 (_mm_srl_epi16): Likewise.
2066 (_mm_srl_epi32): Likewise.
2067 (_mm_srl_epi64): Likewise.
2069 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
2070 Jinsong Ji <jji@us.ibm.com>
2072 * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
2073 * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
2074 __vector __bool int. Use vec_cmpgt in preference to deprecated
2075 function vec_vcmpgtfp.
2076 (_mm_max_ps): Likewise.
2078 2018-10-25 Jeff Law <law@redhat.com>
2080 * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
2083 2018-10-25 Martin Sebor <msebor@redhat.com>
2085 * doc/extend.texi (aligned): Expand attribute description.
2086 (Alignment): Rename section. Discuss function arguments.
2088 2018-10-25 Jan Hubicka <jh@suse.cz>
2090 * ipa-devirt.c (main_odr_variant): Remove.
2091 (hash_odr_name, types_same_for_odr, types_odr_comparable,
2092 odr_name_hasher::equal, odr_subtypes_equivalent_p):
2093 Drop use of main_odr_variant.
2094 (add_type_duplicate): Silence confused warnings on integer types.
2095 (get_odr_type): Always look for main variant.
2096 (register_odr_type): Simplify.
2098 2018-10-25 Richard Biener <rguenther@suse.de>
2100 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
2101 Initialize ng to silence error with release checking bootstrap.
2103 2018-10-25 Richard Biener <rguenther@suse.de>
2105 * tree-if-conv.c: Include tree-ssa-sccvn.h.
2106 (tree_if_conversion): Run CSE on the if-converted loop body.
2108 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
2110 * config/s390/constraints.md (ZL): New constraint.
2111 * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
2113 * config/s390/s390.md (movdi_larl): Remove.
2114 (movdi_64): Add the LARL alternative.
2116 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
2119 * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
2120 (RTX_CODE_HWINT_P): New macro.
2121 (rtx_code_size): Use RTX_CODE_HWINT_P ().
2123 2018-10-25 Jan Hubicka <jh@suse.cz>
2125 * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
2128 2018-10-25 Richard Biener <rguenther@suse.de>
2130 PR tree-optimization/87665
2131 PR tree-optimization/87745
2132 * tree-vectorizer.h (get_earlier_stmt): Remove.
2133 (get_later_stmt): Pick up UID from the original non-pattern stmt.
2135 2018-10-25 Sam Tebbs <sam.tebbs@arm.com>
2137 * options.texi (Deprecated): Move list to Var section.
2139 2018-10-24 Bill Schmidt <wschmidt@linux.ibm.com>
2140 Jinsong Ji <jji@us.ibm.com>
2142 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
2143 __vector long to __vector long long.
2144 (_mm_cvtpd_ps): Likewise.
2145 (_mm_cvttpd_epi32): Likewise.
2146 (_mm_cvtpi32_pd): Likewise.
2147 (_mm_unpackhi_epi64): Likewise.
2148 (_mm_unpacklo_epi64): Likewise.
2150 2018-10-24 Segher Boessenkool <segher@kernel.crashing.org>
2152 PR rtl-optimization/87720
2153 * combine.c (make_more_copies): Skip if the dest is pc_rtx.
2155 2018-10-24 Alexandre Oliva <aoliva@redhat.com>
2157 * gimple-ssa-isolate-paths.c
2158 (find_implicit_erroneous_behavior): Do not change code if the
2159 pass is running for warnings only.
2160 (find_explicit_erroneous_behavior): Likewise.
2162 2018-10-24 Michael Meissner <meissner@linux.ibm.com>
2164 * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
2165 Define as rs6000_mangle_decl_assembler_name.
2166 (rs6000_mangle_decl_assembler_name): If the user switched from IBM
2167 long double to IEEE long double, switch the names of the long
2168 double built-in functions to be <func>f128 instead of <func>l.
2170 2018-10-24 Martin Sebor <msebor@redhat.com>
2172 * doc/extend.texi (nonnull): List no-argument form. Reference
2173 -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
2175 2018-10-24 Richard Biener <rguenther@suse.de>
2177 * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
2179 2018-10-24 Martin Liska <mliska@suse.cz>
2181 PR tree-optimization/84436
2182 * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
2184 (switch_conversion::contains_linear_function_p): New.
2185 (switch_conversion::build_one_array): Support linear
2186 transformation on input.
2187 * tree-switch-conversion.h (struct switch_conversion): Add
2188 contains_linear_function_p declaration.
2190 2018-10-24 Richard Biener <rguenther@suse.de>
2192 * varasm.c (const_hash_1): Return hash of ADDR_EXPR
2193 if its argument is CONSTANT_CLASS_P.
2195 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
2197 * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
2198 it is wrong for forward declarations.
2200 2018-10-24 Ilya Leoshkevich <iii@linux.ibm.com>
2202 * config/s390/s390.c (s390_check_qrst_address): Add the missing
2203 SYMBOL_REF_P () check.
2205 2018-10-24 Richard Biener <rguenther@suse.de>
2207 PR tree-optimization/87105
2208 * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
2209 dump classification.
2210 (vect_analyze_data_ref_accesses): Handle duplicate loads and
2211 stores by splitting the affected group after the fact.
2212 * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
2213 fail the SLP build because of size constraints.
2215 2018-10-24 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2217 * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
2218 * configure: Regenerate.
2219 * config.in: Regenerate.
2220 * varasm.c (mergeable_string_section): Use readonly_data_section
2221 if linker doesn't support SHF_MERGE with alignment > 8.
2222 (mergeable_constant_section): Likewise.
2224 2018-10-24 Richard Biener <rguenther@suse.de>
2226 PR tree-optimization/84013
2227 * tree-ssa-structalias.c (struct msdi_data): New struct for
2228 marshalling data to walk_stmt_load_store_ops.
2229 (maybe_set_dependence_info): Refactor as callback for
2230 walk_stmt_load_store_ops.
2231 (compute_dependence_clique): Set restrict info on all stmt kinds.
2233 2018-10-24 Martin Liska <mliska@suse.cz>
2235 * cgraph.c (cgraph_node::dump):
2236 Remove reduntant dumps and make tp_first_run dump more compact.
2238 2018-10-24 Richard Biener <rguenther@suse.de>
2240 PR tree-optimization/87665
2241 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
2244 2018-10-12 Jeff Law <law@redhat.com>
2246 * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
2249 2018-10-23 Richard Biener <rguenther@suse.de>
2251 * tree-vrp.c (add_assert_info): Guard dump_printf with
2253 * gimple-ssa-evrp-analyze.c
2254 (evrp_range_analyzer::record_ranges_from_incoming_edge):
2255 Use value_range::ignore_equivs_equal_p.
2257 2018-10-23 Richard Biener <rguenther@suse.de>
2259 PR tree-optimization/87105
2260 PR tree-optimization/87608
2261 * passes.def (pass_all_early_optimizations): Add early phi-opt
2263 * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
2264 addition to debug stmts.
2265 (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
2266 and abs replacement early.
2267 * tree-cfg.c (gimple_empty_block_p): Likewise.
2269 2018-10-23 Richard Earnshaw <rearnsha@arm.com>
2272 * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
2273 specified to configure.
2274 (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
2276 2018-10-23 Richard Biener <rguenther@suse.de>
2278 PR tree-optimization/87700
2279 * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
2281 2018-10-23 Jakub Jelinek <jakub@redhat.com>
2284 * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
2285 second argument from __mmask16 to __mmask8.
2286 * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
2287 _mm_mask_packs_epi32): Likewise.
2288 * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
2290 (_mm512_mask_prefetch_i64scatter_pd): Likewise. Formatting fix.
2292 2018-10-23 Richard Biener <rguenther@suse.de>
2294 * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
2296 2018-10-23 Richard Biener <rguenther@suse.de>
2298 PR tree-optimization/86144
2299 * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
2300 over simd attribute.
2302 2018-10-23 Richard Biener <rguenther@suse.de>
2304 PR tree-optimization/87693
2305 * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
2306 the case we do not find the taken edge.
2308 2018-10-22 Bill Schmidt <wschmidt@linux.ibm.com>
2309 Jinsong Ji <jji@us.ibm.com>
2311 * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
2312 (_mm_store_pd): Use unaligned vector type for pointer cast.
2313 (_mm_maskmoveu_si128): Likewise.
2314 * config/rs6000/xmmintrin.h (__m128_u): New typedef.
2315 (_mm_store_ps): Use unaligned vector type for pointer cast.
2317 2018-10-22 Paul Koning <ni1d@arrl.net>
2319 * symtab.c (symtab_node::increase_alignment): Correct max
2322 2018-10-22 Yury Gribov <tetra2005@gmail.com>
2324 PR tree-optimization/87633
2325 * match.pd: Do not generate unordered integer comparisons.
2327 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
2329 PR rtl-optimization/87600
2330 * combine.c: Add include of expr.h.
2331 (cant_combine_insn_p): Do not combine moves from any hard non-fixed
2332 register to a pseudo.
2333 (make_more_copies): New function, add a copy to a new pseudo after
2334 the moves from hard registers into pseudos.
2335 (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
2336 later. Call make_more_copies.
2338 2018-10-22 Andrew Stubbs <ams@codesourcery.com>
2340 * lra-constraints.c (process_alt_operands): New local array,
2341 matching_early_clobber. Check matching_early_clobber before
2342 decrementing reject, and set matching_early_clobber after.
2344 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
2347 * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
2348 call output_addr_const and hope for the best.
2350 2018-10-22 Richard Biener <rguenther@suse.de>
2352 * gimple-ssa-evrp-analyze.c
2353 (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
2354 smarter about what ranges to use.
2355 * tree-vrp.c (add_assert_info): Dump here.
2356 (register_edge_assert_for_2): Instead of here at multiple but
2359 * gcc.dg/tree-ssa/evrp12.c: New testcase.
2360 * gcc.dg/predict-6.c: Adjust.
2361 * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
2362 * gcc.dg/tree-ssa/vrp02.c: Likewise.
2363 * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
2365 2018-10-22 Steven Bosscher <steven@gcc.gnu.org>
2366 Richard Biener <rguenther@suse.de>
2368 * bitmap.h: Update data structure documentation, including a
2369 description of bitmap views as either linked-lists or splay trees.
2370 (struct bitmap_element_def): Update comments for splay tree bitmaps.
2371 (struct bitmap_head_def): Likewise.
2372 (bitmap_list_view, bitmap_tree_view): New prototypes.
2373 (bitmap_initialize_stat): Initialize a bitmap_head's indx and
2375 (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
2376 (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
2377 * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
2378 released bitmap element here.
2379 (bitmap_element_free): Remove.
2380 (bitmap_elt_clear_from): Work on splay tree bitmaps.
2381 (bitmap_list_link_element): Renamed from bitmap_element_link. Move
2382 this function similar ones such that linked-list bitmap implementation
2383 functions are grouped.
2384 (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
2385 and moved for grouping.
2386 (bitmap_list_insert_element_after): Renamed from
2387 bitmap_elt_insert_after, and moved for grouping.
2388 (bitmap_list_find_element): New function spliced from bitmap_find_bit.
2389 (bitmap_tree_link_left, bitmap_tree_link_right,
2390 bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
2391 bitmap_tree_link_element, bitmap_tree_unlink_element,
2392 bitmap_tree_find_element): New functions for splay-tree bitmap
2394 (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
2395 Renamed and moved, see above entries.
2396 (bitmap_tree_listify_from): New function to convert part of a splay
2397 tree bitmap to a linked-list bitmap.
2398 (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
2399 (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
2400 (bitmap_find_bit): Remove.
2401 (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
2402 bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
2403 Handle splay tree bitmaps.
2404 (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
2405 bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
2406 bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
2407 bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
2408 bitmap_intersect_compl_p, bitmap_ior_and_compl,
2409 bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
2410 bitmap_hash): Reject trying to act on splay tree bitmaps. Make
2411 corresponding changes to use linked-list specific bitmap_element
2412 manipulation functions as applicable for efficiency.
2413 (bitmap_tree_to_vec): New function.
2414 (debug_bitmap_elt_file): New function split out from ...
2415 (debug_bitmap_file): ... here. Handle splay tree bitmaps.
2416 (bitmap_print): Likewise.
2418 PR tree-optimization/63155
2419 * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
2421 * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
2424 2018-10-22 Martin Liska <mliska@suse.cz>
2426 PR tree-optimization/87686
2428 2018-08-29 Martin Liska <mliska@suse.cz>
2430 * tree-switch-conversion.c (switch_conversion::expand):
2431 Strenghten assumption about gswitch statements.
2433 2018-10-22 Martin Liska <mliska@suse.cz>
2435 * ipa-icf.c (sem_item::compare_attributes): Remove.
2436 (sem_item::compare_referenced_symbol_properties): Use
2437 attribute_list_equal instead.
2438 (sem_function::equals_wpa): Likewise.
2439 * ipa-icf.h: Remove compare_attributes.
2441 2018-10-22 Richard Biener <rguenther@suse.de>
2444 * mem-stats.h (mem_usage::operator==): Fix pasto.
2446 2018-10-22 Richard Biener <rguenther@suse.de>
2448 PR tree-optimization/87640
2449 * tree-vrp.c (set_value_range_with_overflow): Decompose
2451 (extract_range_from_binary_expr_1): Adjust.
2453 2018-10-22 Martin Jambor <mjambor@suse.cz>
2455 * tree-eh.h (stmt_could_throw_p): Add function parameter.
2456 (stmt_can_throw_external): Likewise.
2457 (stmt_can_throw_internal): Likewise.
2458 * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
2459 (lower_eh_constructs_2): Likewise.
2460 (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
2461 (stmt_can_throw_external): Likewise.
2462 (stmt_can_throw_internal): Likewise.
2463 (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
2464 (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
2465 (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
2466 (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
2467 (pass_lower_eh_dispatch::execute): Pass cfun to
2468 stmt_can_throw_external.
2469 (cleanup_empty_eh): Likewise.
2470 (verify_eh_edges): Pass cfun to stmt_could_throw_p.
2471 * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
2472 stmt_can_throw_external instead of pushing it to cfun.
2473 (symbol_table::create_edge): Likewise.
2474 * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
2475 stmt_can_throw_internal.
2476 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
2477 to stmt_could_throw_p.
2478 * gimple-ssa-store-merging.c (handled_load): Pass cfun to
2479 stmt_can_throw_internal.
2480 (pass_store_merging::execute): Likewise.
2481 * gimple-ssa-strength-reduction.c
2482 (find_candidates_dom_walker::before_dom_children): Pass cfun to
2484 * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
2485 stmt_can_throw_internal.
2486 * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
2487 to stmt_can_throw_external.
2488 (check_stmt): Pass cfun to stmt_could_throw_p.
2489 (check_stmt): Pass cfun to stmt_can_throw_external.
2490 (pass_nothrow::execute): Likewise.
2491 * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
2492 * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
2493 stmt_can_throw_internal.
2494 (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
2495 (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
2496 (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
2497 * tree-complex.c (expand_complex_libcall): Pass cfun to
2498 stmt_could_throw_p and to stmt_can_throw_internal.
2499 (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
2500 * tree-inline.c (copy_edges_for_bb): Likewise.
2501 (maybe_move_debug_stmts_to_successors): Likewise.
2502 * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
2504 * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
2505 * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
2506 * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
2507 stmt_can_throw_internal.
2508 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
2509 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
2511 (mark_aliased_reaching_defs_necessary_1): Pass cfun to
2512 stmt_can_throw_internal.
2513 * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
2514 * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
2516 * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
2517 (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
2518 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
2519 (convert_mult_to_fma_1): Likewise.
2520 (convert_to_divmod): Likewise.
2521 * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
2522 * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
2523 * tree-ssa-propagate.c
2524 (substitute_and_fold_dom_walker::before_dom_children): Likewise.
2525 * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
2526 (maybe_optimize_range_tests): Likewise.
2527 (linearize_expr_tree): Likewise.
2528 (reassociate_bb): Likewise.
2529 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2530 * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
2531 * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
2532 (handle_char_store): Likewise.
2533 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
2534 stmt_can_throw_internal.
2535 * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
2537 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2538 (vectorizable_call): Pass cfun to stmt_can_throw_internal.
2539 (vectorizable_simd_clone_call): Likewise.
2540 * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
2541 (gimple_stringop_fixed_value): Likewise.
2543 2018-10-22 Ilya Leoshkevich <iii@linux.ibm.com>
2545 * config/s390/s390.c (s390_loadrelative_operand_p): Accept
2546 literal pool references.
2547 (s390_check_qrst_address): Adapt to the new behavior of
2548 s390_loadrelative_operand_p ().
2550 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
2553 * config/i386/sse.md (*andnot<mode>3_bcst): New.
2555 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
2558 * config/i386/sse.md (*<code><mode>3_bcst): New.
2560 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
2563 * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
2564 V4DI, V16SI and V8DI.
2565 (*sub<mode>3<mask_name>_bcst): New.
2566 (*add<mode>3<mask_name>_bcst): Likewise.
2568 2018-10-21 Bill Schmidt <wschmidt@linux.ibm.com>
2569 Jinsong Ji <jji@us.ibm.com>
2571 * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
2572 __m64 with __vector unsigned long long for compatibility.
2573 (_mm_movemask_epi8): Likewise.
2574 * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
2575 (_mm_cvttps_pi32): Likewise.
2576 (_mm_cvtpi32_ps): Likewise.
2577 (_mm_cvtps_pi16): Likewise.
2578 (_mm_loadh_pi): Likewise.
2579 (_mm_storeh_pi): Likewise.
2580 (_mm_movehl_ps): Likewise.
2581 (_mm_movelh_ps): Likewise.
2582 (_mm_loadl_pi): Likewise.
2583 (_mm_storel_pi): Likewise.
2584 (_mm_movemask_ps): Likewise.
2585 (_mm_shuffle_pi16): Likewise.
2587 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
2590 * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
2591 __builtin_ia32_vfnmsubpd512_mask.
2592 (_mm512_mask_fnmsub_round_pd): Likewise.
2593 (_mm512_fnmsub_pd): Likewise.
2594 (_mm512_mask_fnmsub_pd): Likewise.
2595 (_mm512_maskz_fnmsub_round_pd): Use
2596 __builtin_ia32_vfnmsubpd512_maskz.
2597 (_mm512_maskz_fnmsub_pd): Likewise.
2598 (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
2599 (_mm512_mask_fnmsub_round_ps): Likewise.
2600 (_mm512_fnmsub_ps): Likewise.
2601 (_mm512_mask_fnmsub_ps): Likewise.
2602 (_mm512_maskz_fnmsub_round_ps): Use
2603 __builtin_ia32_vfnmsubps512_maskz.
2604 (_mm512_maskz_fnmsub_ps): Likewise.
2605 * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
2606 __builtin_ia32_vfnmsubpd256_mask.
2607 (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
2608 (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
2609 (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
2610 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
2611 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
2612 (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
2613 (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
2614 (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
2615 * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
2616 __builtin_ia32_vfnmsubpd.
2617 (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
2618 (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
2619 (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
2620 (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
2621 (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
2622 * config/i386/i386-builtin.def: Add
2623 __builtin_ia32_vfnmsubpd256_mask,
2624 __builtin_ia32_vfnmsubpd256_maskz,
2625 __builtin_ia32_vfnmsubpd128_mask,
2626 __builtin_ia32_vfnmsubpd128_maskz,
2627 __builtin_ia32_vfnmsubps256_mask,
2628 __builtin_ia32_vfnmsubps256_maskz,
2629 __builtin_ia32_vfnmsubps128_mask,
2630 __builtin_ia32_vfnmsubps128_maskz,
2631 __builtin_ia32_vfnmsubpd512_mask,
2632 __builtin_ia32_vfnmsubpd512_maskz,
2633 __builtin_ia32_vfnmsubps512_mask,
2634 __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
2635 __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
2636 __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
2637 __builtin_ia32_vfnmsubpd256.
2638 * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
2639 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
2640 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
2642 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
2644 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
2646 (fmai_vmfnmsub_<mode><round_name>): Likewise.
2648 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
2651 * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
2652 __builtin_ia32_vfnmaddpd512_mask.
2653 (_mm512_mask_fnmadd_round_pd): Likewise.
2654 (_mm512_fnmadd_pd): Likewise.
2655 (_mm512_mask_fnmadd_pd): Likewise.
2656 (_mm512_maskz_fnmadd_round_pd): Use
2657 __builtin_ia32_vfnmaddpd512_maskz.
2658 (_mm512_maskz_fnmadd_pd): Likewise.
2659 (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
2660 (_mm512_mask_fnmadd_round_ps): Likewise.
2661 (_mm512_fnmadd_ps): Likewise.
2662 (_mm512_mask_fnmadd_ps): Likewise.
2663 (_mm512_maskz_fnmadd_round_ps): Use
2664 __builtin_ia32_vfnmaddps512_maskz.
2665 (_mm512_maskz_fnmadd_ps): Likewise.
2666 * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
2667 __builtin_ia32_vfnmaddpd256_mask.
2668 (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
2669 (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
2670 (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
2671 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
2672 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
2673 (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
2674 (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
2675 (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
2676 * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
2677 __builtin_ia32_vfnmaddpd.
2678 (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
2679 (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
2680 (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
2681 (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
2682 (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
2683 * config/i386/i386-builtin.def: Add
2684 __builtin_ia32_vfnmaddpd256_mask,
2685 __builtin_ia32_vfnmaddpd256_maskz,
2686 __builtin_ia32_vfnmaddpd128_mask,
2687 __builtin_ia32_vfnmaddpd128_maskz,
2688 __builtin_ia32_vfnmaddps256_mask,
2689 __builtin_ia32_vfnmaddps256_maskz,
2690 __builtin_ia32_vfnmaddps128_mask,
2691 __builtin_ia32_vfnmaddps128_maskz,
2692 __builtin_ia32_vfnmaddpd512_mask,
2693 __builtin_ia32_vfnmaddpd512_maskz,
2694 __builtin_ia32_vfnmaddps512_mask,
2695 __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
2696 __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
2697 __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
2698 __builtin_ia32_vfnmaddpd256.
2699 * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
2700 (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
2701 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
2703 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
2705 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
2707 (fmai_vmfnmadd_<mode><round_name>): Likewise.
2709 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
2712 * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
2713 __builtin_ia32_vfmsubpd512_mask.
2714 (_mm512_mask_fmsub_round_pd): Likewise.
2715 (_mm512_fmsub_pd): Likewise.
2716 (_mm512_mask_fmsub_pd): Likewise.
2717 (_mm512_maskz_fmsub_round_pd): Use
2718 __builtin_ia32_vfmsubpd512_maskz.
2719 (_mm512_maskz_fmsub_pd): Likewise.
2720 (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
2721 (_mm512_mask_fmsub_round_ps): Likewise.
2722 (_mm512_fmsub_ps): Likewise.
2723 (_mm512_mask_fmsub_ps): Likewise.
2724 (_mm512_maskz_fmsub_round_ps): Use
2725 __builtin_ia32_vfmsubps512_maskz.
2726 (_mm512_maskz_fmsub_ps): Likewise.
2727 * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
2728 __builtin_ia32_vfmsubpd256_mask.
2729 (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
2730 (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
2731 (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
2732 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
2733 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
2734 (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
2735 (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
2736 (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
2737 * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
2738 __builtin_ia32_vfmsubpd.
2739 (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
2740 (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
2741 (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
2742 (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
2743 (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
2744 * config/i386/i386-builtin.def: Add
2745 __builtin_ia32_vfmsubpd256_mask,
2746 __builtin_ia32_vfmsubpd256_maskz,
2747 __builtin_ia32_vfmsubpd128_mask,
2748 __builtin_ia32_vfmsubpd128_maskz,
2749 __builtin_ia32_vfmsubps256_mask,
2750 __builtin_ia32_vfmsubps256_maskz,
2751 __builtin_ia32_vfmsubps128_mask,
2752 __builtin_ia32_vfmsubps128_maskz,
2753 __builtin_ia32_vfmsubpd512_mask,
2754 __builtin_ia32_vfmsubpd512_maskz,
2755 __builtin_ia32_vfmsubps512_mask,
2756 __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
2757 __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
2758 __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
2759 __builtin_ia32_vfmsubpd256.
2760 * config/i386/sse.md (fma4i_fmsub_<mode>): New.
2761 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
2762 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
2764 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
2766 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
2768 (fmai_vmfmsub_<mode><round_name>): Likewise.
2770 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
2772 * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
2773 Remove plus. Renamed to ...
2774 (*sub<mode>3<mask_name>_bcst): This.
2775 (*add<mode>3<mask_name>_bcst_2): Renamede to ...
2776 (*add<mode>3<mask_name>_bcst): This.
2778 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
2781 * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
2783 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
2786 * i386/avx512vlintrin.h (_mm256_or_epi32): New.
2787 (_mm_or_epi32): Likewise.
2788 (_mm256_xor_epi32): Likewise.
2789 (_mm_xor_epi32): Likewise.
2790 (_mm256_or_epi64): Likewise.
2791 (_mm_or_epi64): Likewise.
2792 (_mm256_xor_epi64): Likewise.
2793 (_mm_xor_epi64): Likewise.
2795 2018-10-20 H.J. Lu <hongjiu.lu@intel.com>
2798 * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
2800 2018-10-20 Jakub Jelinek <jakub@redhat.com>
2803 * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
2805 2018-10-20 Andreas Schwab <schwab@linux-m68k.org>
2807 * doc/ux.texi: Move @section directly after @node.
2809 2018-10-19 Jakub Jelinek <jakub@redhat.com>
2813 * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
2814 depend closely nested inside of loop with ordered clause with
2817 2018-10-19 David Malcolm <dmalcolm@redhat.com>
2819 * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
2820 * doc/gccint.texi: Include ux.texi and use it in top-level menu.
2821 * doc/ux.texi: New file.
2823 2018-10-19 Segher Boessenkool <segher@kernel.crashing.org>
2825 * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
2826 be the first CR field allocated.
2828 2018-10-19 Richard Biener <rguenther@suse.de>
2831 * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
2832 TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
2834 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
2837 * config/i386/sse.md
2838 (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
2839 (*add<mode>3<mask_name>_bcst_2): Likewise.
2841 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
2843 * config/i386/sse.md
2844 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
2845 Replace nonimmediate_operand with register_operand.
2846 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
2848 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
2851 2018-10-19 Ilya Leoshkevich <iii@linux.ibm.com>
2853 PR rtl-optimization/87596
2854 * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
2855 lra_get_insn_recog_data () instead of lra_insn_recog_data[]
2856 for instructions in FROM..TO range.
2858 2018-10-19 Eric Botcazou <ebotcazou@adacore.com>
2860 * cfgexpand.c (expand_one_var): Use specific wording in error message
2861 for non-local frame variables.
2862 * stor-layout.c (layout_decl): Do not issue a warning for them.
2864 2018-10-19 Robin Dapp <rdapp@linux.ibm.com>
2866 * haifa-sched.c (priority): Add force_recompute parameter.
2867 (apply_replacement): Call priority () with force_recompute = true.
2868 (restore_pattern): Likewise.
2870 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
2872 * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
2873 HOST_BITS_PER_WIDE_INT.
2874 (test_vector_ops_duplicate): Likewise.
2876 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
2879 * config/i386/sse.md (VF_AVX512): New.
2880 (avx512bcst): Likewise.
2881 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
2883 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
2885 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
2888 2018-10-18 Jonathan Wakely <jwakely@redhat.com>
2890 * doc/invoke.texi (-dumpversion): Improve grammar.
2891 (-dumpfullversion): Make more consistent with -dumpversion.
2893 2018-10-18 Uros Bizjak <ubizjak@gmail.com>
2895 * config/i386/i386.c (ix86_emit_fp_unordered_jump):
2896 Set JUMP_LABEL to the jump insn.
2897 (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
2898 Predict emitted jump and add label to jump insn.
2900 2018-10-18 David Malcolm <dmalcolm@redhat.com>
2902 PR tree-optimization/87562
2903 * input.c (get_substring_ranges_for_loc): Use
2904 LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
2905 getting the linemap for the endpoint. Verify that it's either
2906 in the same linemap as the start point's spelling location, or
2907 at least in the same file.
2909 2018-10-18 Richard Biener <rguenther@suse.de>
2911 * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
2912 feed width-specific load/store costs through ix86_vec_cost.
2913 * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
2914 (k8_cost): Likewise.
2915 (bdver_cost): Likewise.
2916 (znver1_cost): Likewise.
2917 (btver1_cost): Likewise.
2918 (btver2_cost): Likewise.
2920 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
2922 * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
2923 to simplify subreg of vec_merge.
2925 2018-10-18 Richard Biener <rguenther@suse.de>
2927 * config/i386/i386.c: Fix costing of vector FMA.
2929 2018-10-18 Richard Biener <rguenther@suse.de>
2931 * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
2933 (ix86_builtin_vectorization_cost): For vec_construct properly
2934 cost insertion into SSE regs.
2935 (...): Adjust calls to ix86_vec_cost.
2937 2018-10-18 Richard Biener <rguenther@suse.de>
2941 2018-02-07 Richard Biener <rguenther@suse.de>
2943 PR tree-optimization/84204
2944 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2947 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
2950 * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
2952 (test_vector_ops_duplicate): Add test for a scalar subreg of a
2953 VEC_MERGE of a VEC_DUPLICATE.
2955 2018-10-17 Joseph Myers <joseph@codesourcery.com>
2957 * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
2958 * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
2959 * doc/standards.texi (C Language): Document C2X.
2960 * dwarf2out.c (highest_c_language), config/rl78/rl78.c
2961 (rl78_option_override): Handle "GNU C2X" language name.
2963 2018-10-17 Joseph Myers <joseph@codesourcery.com>
2965 * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
2966 Document C17 as published in 2018.
2968 2018-10-17 Eric Botcazou <ebotcazou@adacore.com>
2971 * fold-const.c (fold_truth_andor_1): If the right side is not constant,
2972 bail out if both sides do not have the same storage order.
2974 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
2976 * bitmap.c (bitmap_head::dump): New.
2977 * bitmap.h (bitmap_head): Add dump().
2978 * gimple-ssa-evrp-analyze.c
2979 (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
2980 (evrp_range_analyzer::set_ssa_range_info): Same.
2981 (evrp_range_analyzer::record_ranges_from_phis): Same.
2982 (evrp_range_analyzer::record_ranges_from_stmt): Same.
2983 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
2984 * gimple-ssa-sprintf.c (get_int_range): Same.
2985 (format_integer): Same.
2986 (sprintf_dom_walker::handle_gimple_call): Same.
2987 * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
2988 (ipcp_vr_lattice::top_p): Same.
2989 (ipcp_vr_lattice::bottom_p): Same.
2990 (ipcp_vr_lattice::set_to_bottom): Same.
2991 (ipa_vr_operation_and_type_effects): Same.
2992 (propagate_vr_across_jump_function): Same.
2993 (ipcp_store_vr_results): Same.
2994 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
2995 (ipa_print_node_jump_functions_for_edge): Same.
2996 (ipa_get_value_range): Same.
2997 (ipa_compute_jump_functions_for_edge): Same.
2998 (ipa_write_jump_function): Same.
2999 * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
3000 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
3002 * vr-values.c (set_value_range_to_nonnegative): Same.
3003 (set_value_range_to_truthvalue): Same.
3004 (vr_values::get_value_range): Same.
3005 (vr_values::set_defs_to_varying): Same.
3006 (vr_values::update_value_range): Same.
3007 (symbolic_range_based_on_p): Same.
3008 (vr_values::op_with_boolean_value_range_p): Same.
3009 (vr_values::extract_range_for_var_from_comparison_expr): Same.
3010 (vr_values::extract_range_from_ssa_name): Same.
3011 (vr_values::extract_range_from_binary_expr): Same.
3012 (vr_values::extract_range_from_unary_expr): Same.
3013 (vr_values::extract_range_from_cond_expr): Same.
3014 (vr_values::extract_range_from_comparison): Same.
3015 (vr_values::check_for_binary_op_overflow): Same.
3016 (vr_values::extract_range_basic): Same.
3017 (vr_values::extract_range_from_assignment): Same.
3018 (compare_ranges): Same.
3019 (compare_range_with_value): Same.
3020 (vr_values::adjust_range_with_scev): Same.
3021 (vrp_valueize): Same.
3022 (vrp_valueize_1): Same.
3023 (vr_values::get_vr_for_comparison): Same.
3024 (vr_values::compare_name_with_value): Same.
3025 (vr_values::compare_names): Same.
3026 (vr_values::vrp_evaluate_conditional): Same.
3027 (find_case_label_ranges): Same.
3028 (vr_values::vrp_visit_switch_stmt): Same.
3029 (vr_values::extract_range_from_phi_node): Same.
3030 (vr_values::simplify_div_or_mod_using_ranges): Same.
3031 (vr_values::simplify_bit_ops_using_ranges): Same.
3032 (test_for_singularity): Same.
3033 (range_fits_type_p): Same.
3034 (vr_values::simplify_cond_using_ranges_1): Same.
3035 (vr_values::simplify_switch_using_ranges): Same.
3036 (vr_values::simplify_float_conversion_using_ranges): Same.
3037 (vr_values::two_valued_val_range_p): Same.
3038 (vr_values::add_equivalence): Move to value_range::equiv_add.
3039 * vr-values.h (vr_values::add_equivalence): Remove.
3040 (VR_INITIALIZER): Remove.
3041 * tree-vrp.c (value_range::set): New.
3042 (value_range::equiv_add): New.
3043 (value_range::value_range): New.
3044 (value_range::deep_copy): New.
3045 (value_range::check): New.
3046 (value_range::equal_p): New.
3047 (value_range::ignore_equivs_equal_p): New.
3048 (value_range::operator==): New.
3049 (value_range::operator!=): New.
3050 (value_range::symbolic_p): New.
3051 (value_range::numeric_p): New.
3052 (value_range::set_undefined): New.
3053 (value_range::set_varying): New.
3054 (value_range::may_contain_p): New.
3055 (value_range::equiv_clear): New.
3056 (value_range::singleton_p): New.
3057 (value_range::intersect): New.
3058 (value_range::dump): New.
3059 (value_range::set_and_canonicalize): New.
3060 (set_value_range): Adjust for value_range API.
3061 (set_value_range_to_undefined): Same.
3062 (set_value_range_to_varying): Same.
3063 (set_and_canonicalize_value_range): Same.
3064 (set_value_range_to_nonnull): Same.
3065 (set_value_range_to_null): Same.
3066 (range_is_null): Same.
3067 (range_is_nonnull): Same.
3068 (range_int_cst_p): Same.
3069 (range_int_cst_singleton_p): Same.
3070 (symbolic_range_p): Same.
3071 (range_includes_zero_p): Same.
3072 (value_range_constant_singleton): Same.
3073 (vrp_set_zero_nonzero_bits): Same.
3074 (ranges_from_anti_range): Same.
3075 (extract_range_into_wide_ints): Same.
3076 (extract_range_from_multiplicative_op): Same.
3077 (set_value_range_with_overflow): Same.
3078 (extract_range_from_binary_expr_1): Same.
3079 (extract_range_from_unary_expr): Same.
3080 (dump_value_range): Same.
3081 (debug_value_range): Same.
3082 (vrp_prop::check_array_ref): Same.
3083 (vrp_prop::check_mem_ref): Same.
3084 (vrp_prop::vrp_initialize): Same.
3085 (vrp_prop::visit_stmt): Same.
3086 (intersect_ranges): Same.
3087 (vrp_prop::visit_phi): Same.
3088 (vrp_prop::vrp_finalize): Same.
3089 (determine_value_range_1): Same.
3090 (determine_value_range): Same.
3091 (vrp_intersect_ranges_1): Rename to...
3092 (vrp_intersect_1): this.
3093 (vrp_intersect_ranges): Rename to...
3094 (value_range::intersect_helper): ...this.
3095 (vrp_meet_1): Rename to...
3096 (value_range::union_helper): ...this.
3097 (vrp_meet): Rename to...
3098 (value_range::union_): ...this.
3099 (copy_value_range): Remove.
3100 * tree-vrp.h (struct value_range): Rewrite into a proper class.
3101 (value_range::vrtype): New.
3102 (value_range::type): New.
3103 (value_range::equiv): New.
3104 (value_range::min): New.
3105 (value_range::max): New.
3106 (value_range::varying_p): New.
3107 (value_range::undefined_p): New.
3108 (value_range::null_p): New.
3109 (value_range::equiv_add): New.
3110 (copy_value_range): Remove.
3112 2018-10-17 David Malcolm <dmalcolm@redhat.com>
3114 * Makefile.in (SELFTEST_TARGETS): New.
3115 (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
3116 (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
3117 (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
3119 (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
3120 (selftest-c++-gdb, selftest-c++-valgrind): Move to
3122 * configure: Regenerate.
3123 * configure.ac (selftest_languages): New.
3125 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
3127 * tree-vrp.c (extract_range_from_multiplicative_op): Remove
3128 overflow wraps argument.
3129 (extract_range_from_binary_expr_1): Do not pass overflow wraps to
3130 wide_int_range_multiplicative_op.
3131 * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
3132 overflow wraps argument.
3133 (wide_int_range_multiplicative_op): Same.
3134 (wide_int_range_lshift): Same.
3135 (wide_int_range_div): Same.
3136 * wide-int-range.h (wide_int_range_multiplicative_op): Same.
3137 (wide_int_range_lshift): Same.
3138 (wide_int_range_div): Same.
3140 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
3142 * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
3143 use sign as argument.
3144 * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
3145 wide_int_range_shift_undefined_p.
3147 2018-10-16 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3149 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
3151 (@despeculate_copy<ALLI_TI:mode>): ... This.
3152 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
3155 2018-10-16 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3157 * config.gcc: Obsolete *-*-solaris2.10*.
3158 * doc/install.texi (Specific, *-*-solaris2*): Document it.
3160 2018-10-12 Jeff Law <law@redhat.com>
3162 * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
3163 reg + sym +- const_int addressing modes.
3165 2018-10-15 David Malcolm <dmalcolm@redhat.com>
3167 * common.opt (fdiagnostics-minimum-margin-width=): New option.
3168 * diagnostic-show-locus.c (layout::layout): Apply the minimum
3170 (layout::start_annotation_line): Only print up to 3 of the
3171 margin character, to avoid touching the left-hand side.
3172 (selftest::test_diagnostic_show_locus_fixit_lines): Update for
3173 minimum margin width, as set by test_diagnostic_context's ctor.
3174 (selftest::test_fixit_insert_containing_newline): Likewise.
3175 (selftest::test_fixit_insert_containing_newline_2): Likewise.
3176 (selftest::test_line_numbers_multiline_range): Clear
3177 dc.min_margin_width.
3178 * diagnostic.c (diagnostic_initialize): Initialize
3180 * diagnostic.h (struct diagnostic_context): Add field
3182 * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
3183 * opts.c (common_handle_option): Handle
3184 OPT_fdiagnostics_minimum_margin_width_.
3185 * selftest-diagnostic.c
3186 (selftest::test_diagnostic_context::test_diagnostic_context):
3187 Initialize min_margin_width to 6.
3188 * toplev.c (general_init): Initialize global_dc->min_margin_width.
3190 2018-10-15 David Malcolm <dmalcolm@redhat.com>
3192 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3193 Fix usage of "error_at_rich_loc" in the comment.
3195 2018-10-15 Renlin Li <renlin.li@arm.com>
3198 * tree-vectorizer.c (try_vectorize_loop_1): Don't use
3199 if-conversioned loop when it contains ifn with types not
3200 supported by backend.
3201 * internal-fn.c (expand_direct_optab_fn): Add an assert.
3202 (direct_internal_fn_supported_p): New helper function.
3203 * internal-fn.h (direct_internal_fn_supported_p): Declare.
3205 2018-10-15 Jakub Jelinek <jakub@redhat.com>
3208 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
3209 Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
3210 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
3212 2018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com>
3214 PR tree-optimization/87022
3215 * tree-loop-distribution.c (pg_add_dependence_edges): Check all
3216 bits in dist vector rather than the first one.
3218 2018-10-15 Richard Biener <rguenther@suse.de>
3221 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
3222 (visit_loadstore): When a used restrict tag escaped verify that
3223 the points-to solution of "other" pointers do not include
3225 (compute_dependence_clique): If a used restrict tag escaped
3226 communicated that down to visit_loadstore.
3228 2018-10-15 Andreas Krebbel <krebbel@linux.ibm.com>
3230 * config/s390/s390.c (s390_expand_vec_init): Force vector element
3231 into reg if it isn't a general operand.
3233 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
3236 * config/i386/sse.md (*vec_dupv2di): Add register source to
3239 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
3242 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
3243 Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
3244 OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
3245 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
3247 2018-10-13 Eric Botcazou <ebotcazou@adacore.com>
3249 * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
3250 (notice_args_size): Set it in the current trace if no insn that can
3251 throw internally has been seen yet.
3252 (connect_traces): When connecting args_size between traces, allow the
3253 incoming values not to match if there is an insn setting it before the
3254 first insn that can throw internally; in that case, force the creation
3255 of a CFI note on this latter insn.
3257 2018-10-13 Jonathan Wakely <jwakely@redhat.com>
3259 * opt-problem.h (opt_wrapper): Use template-argument-list when naming
3260 the base class, because using the injected-class-name was not clearly
3261 specified until DR 176.
3263 2018-10-12 Paul Koning <ni1d@arrl.net>
3265 * config/pdp11/pdp11.md (doloop_end): New expander.
3266 (doloop_end_insn): renamed from "doloop_end".
3267 (addqi3): New pattern.
3268 (subqi3): New pattern.
3269 * config/pdp11/predicates.md (incdec_operand): New predicate.
3271 2018-10-12 Yury Gribov <tetra2005@gmail.com>
3274 * real.c (format_helper::can_represent_integral_type_p): New function
3275 * real.h (format_helper::can_represent_integral_type_p): Ditto.
3276 * match.pd: New pattern.
3278 2018-10-12 Alexandre Oliva <oliva@adacore.com>
3280 * configure.ac: Introduce --enable-large-address-aware
3281 to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
3282 * doc/install.texi: Document it.
3283 * configure, config.in: Rebuilt.
3284 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
3285 based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
3286 (LINK_SPEC): Insert it.
3287 * config/i386/mingw-w64.h: Likewise.
3289 * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
3291 2018-10-12 Peter Bergner <bergner@linux.ibm.com>
3293 PR rtl-optimization/87600
3294 * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
3296 2018-10-12 Paul Koning <ni1d@arrl.net>
3298 * doc/md.texi (doloop_end): Document that the pattern code may
3299 need to check operand mode.
3301 2018-10-12 Wilco Dijkstra <wdijkstr@arm.com>
3303 * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
3304 to zero-extend between int and floating-point registers.
3305 (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
3306 ldp into floating-point registers. Add type and arch attributes.
3307 (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
3308 Use f_loads for type attribute.
3310 2018-10-11 Martin Sebor <msebor@redhat.com>
3312 * doc/extend.texi (attribute packed): Correct typos.
3314 2018-10-11 Martin Sebor <msebor@redhat.com>
3316 * doc/extend.texi (attribute flatten): Mention interaction with
3319 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
3322 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3323 Set new_decl virtual flag to zero.
3325 2018-10-11 Martin Sebor <msebor@redhat.com>
3328 * doc/extend.texi (attribute format_arg): Discuss using multiple
3329 attributes on a single function.
3331 2018-10-11 Giuliano Belinassi <giuliano.belinassi@usp.br>
3333 PR tree-optimization/86829
3334 * match.pd (sin (atan (x))): New simplification rules.
3335 (cos (atan (x))): Likewise.
3336 * real.c (build_sinatan_real): New function.
3337 * real.h (build_sinatan_real): Prototype.
3339 2018-10-11 Will Schmidt <will_schmidt@vnet.ibm.com>
3341 * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
3343 (fold_mergeeo_helper): New helper function.
3344 (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
3345 intrinsics. Correct some whitespace indentation issues.
3347 2018-10-11 Wilco Dijkstra <wdijkstr@arm.com>
3350 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
3351 Use HOST_WIDE_INT_1U for shift.
3353 2018-10-11 Doug Rupp <rupp@adacore.com>
3354 Olivier Hainque <hainque@adacore.com>
3356 * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
3357 Pass --relax to the linker for RTPs.
3358 (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
3360 2018-10-11 Andrew Stubbs <ams@codesourcery.com>
3361 Jan Hubicka <jh@suse.cz>
3362 Martin Jambor <mjambor@suse.cz>
3364 * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
3365 the same elements are repeated rather than printing all of them.
3366 * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
3367 "repeated" elements.
3368 * read-rtl-function.c (test_loading_repeat): New function.
3369 (read_rtl_function_c_tests): Call test_loading_repeat.
3370 * rtl-tests.c (test_dumping_repeat): New function.
3371 (rtl_tests_c_tests): Call test_dumping_repeat.
3373 2018-10-11 Richard Biener <rguenther@suse.de>
3375 * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
3376 bdver?_cost): Unify to ...
3377 (bdver_memcpy, bdver_memset, bdver_cost): ... this.
3378 * config/i386/i386.c (processor_cost_table): Adjust.
3380 2018-10-10 Eric Botcazou <ebotcazou@adacore.com>
3383 * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
3384 the thunk when expanding to GIMPLE.
3386 2018-10-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
3388 * varasm.c (mergeable_string_section): Don't try to move zero-length
3389 strings to the merge section.
3391 2018-10-10 Uros Bizjak <ubizjak@gmail.com>
3394 * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
3396 2018-10-10 Jakub Jelinek <jakub@redhat.com>
3399 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
3400 to special_args set.
3402 2018-10-10 Richard Biener <rguenther@suse.de>
3404 * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
3405 reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
3406 reduc_plus_scal_v4sf): Merge into pattern reducing to half width
3407 and recursing and pattern terminating the recursion on SSE
3408 vector width using ix86_expand_reduc.
3409 (reduc_sminmax_scal_<mode>): Split into part reducing to half
3410 width and recursing and SSE2 vector variant doing the final
3411 reduction with ix86_expand_reduc.
3412 (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
3413 with terminating the recursion at AVX level, splitting that
3416 2018-10-09 David Malcolm <dmalcolm@redhat.com>
3418 * genmatch.c (error_cb): Rename to...
3419 (diagnostic_cb): ...this, converting int params to enums.
3420 (fatal_at): Update for renaming.
3421 (warning_at): Likewise.
3423 * input.c (selftest::ebcdic_execution_charset::apply):
3424 Update for renaming of...
3425 (selftest::ebcdic_execution_charset::on_error): ...this, renaming
3427 (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
3428 converting level and reason to enums.
3429 (class selftest::lexer_error_sink): Rename to...
3430 (class selftest::lexer_test_options): ...this, renaming field
3431 "m_errors" to "m_diagnostics".
3432 (selftest::lexer_test_options::apply): Update for renaming of...
3433 (selftest::lexer_test_options::on_error): ...this, renaming to...
3434 (selftest::lexer_test_options::on_diagnostic): ...this
3435 converting level and reason to enums.
3436 (selftest::test_lexer_string_locations_raw_string_unterminated):
3437 Update for renamings.
3438 * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
3441 2018-10-09 Paul A. Clarke <pc@us.ibm.com>
3443 * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
3444 * config/rs6000/pmmintrin.h: New file.
3446 2018-10-09 Eric Botcazou <ebotcazou@adacore.com>
3448 PR tree-optimization/86659
3449 * gimple-match.h (gimple_match_op constructors): Initialize reverse.
3451 2018-10-09 Richard Biener <rguenther@suse.de>
3453 PR tree-optimization/63155
3454 * tree-ssa-structalias.c: Include tree-ssa.h.
3455 (get_constraint_for_ssa_var): For undefs return nothing_id.
3456 (find_func_aliases): Cleanup PHI handling.
3458 2018-10-09 Richard Biener <rguenther@suse.de>
3460 * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
3463 2018-10-09 Martin Liska <mliska@suse.cz>
3465 * asan.c (asan_emit_stack_protection): If a stack variable
3466 is located in a same file as current function, then emit
3467 line info into variable definition string.
3469 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
3471 * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
3474 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
3476 * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
3479 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
3481 PR tree-optimization/86659
3482 * gimple-match.h (struct gimple_match_op): Add reverse field.
3483 (gimple_match_op::set_op): New overloaded method.
3484 * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
3485 the REF_REVERSE_STORAGE_ORDER flag on the value.
3486 (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
3487 REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
3489 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
3492 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
3493 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
3495 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
3498 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
3499 Defined with __builtin_ia32_vfmaddsubpd512_mask.
3501 2018-10-08 Richard Biener <rguenther@suse.de>
3503 * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
3504 cost the same as AVX128 ones.
3506 2018-10-08 Paul Koning <ni1d@arrl.net>
3508 * config/pdp11/pdp11-protos.h (output_block_move): Remove.
3509 (expand_block_move): New function.
3510 * config/pdp11/pdp11.c (output_block_move): Remove.
3511 (expand_block_move): New function.
3512 * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
3513 * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
3514 (*movmemhi1): Remove.
3516 2018-10-08 Robin Dapp <rdapp@linux.ibm.com>
3518 * config/s390/2827.md: Increase latencies for some FP instructions.
3520 2018-10-08 Richard Biener <rguenther@suse.de>
3522 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3524 * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
3525 * tree-vectorizer.h (dump_stmt_cost): Adjust.
3526 (add_stmt_cost): Dump return value of the hook.
3528 2018-10-08 Richard Biener <rguenther@suse.de>
3530 PR tree-optimization/63155
3531 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
3532 (ssa_propagation_engine::ssa_propagate): Remove redundant
3533 bitmap bit clearing.
3535 2018-10-05 Peter Bergner <bergner@linux.ibm.com>
3537 PR rtl-optimization/86939
3538 PR rtl-optimization/87479
3539 * ira.h (non_conflicting_reg_copy_p): New prototype.
3540 * ira-lives.c (ignore_reg_for_conflicts): New static variable.
3541 (make_hard_regno_dead): Don't add conflicts for register
3542 ignore_reg_for_conflicts.
3543 (make_object_dead): Likewise.
3544 (non_conflicting_reg_copy_p): New function.
3545 (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
3546 Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
3547 * lra-lives.c (ignore_reg_for_conflicts): New static variable.
3548 (make_hard_regno_dead): Don't add conflicts for register
3549 ignore_reg_for_conflicts. Remove special conflict handling of
3550 REAL_PIC_OFFSET_TABLE_REGNUM. Remove now unused argument
3551 check_pic_pseudo_p and update callers.
3552 (mark_pseudo_dead): Don't add conflicts for register
3553 ignore_reg_for_conflicts.
3554 (process_bb_lives): Set ignore_reg_for_conflicts for copies.
3556 2018-10-05 Andrew Waterman <andrew@sifive.com>
3557 Jim Wilson <jimw@sifive.com>
3559 * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
3560 Add define_expand. Add ! HONOR_SNANS check to current pattern. Add
3561 new pattern using HONOR_SNANS that emits one extra instruction.
3563 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
3565 * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
3566 patterns): Merge SI and DI patterns to a GPR pattern.
3567 (unnamed define_insn and define_split for record form of that): Merge
3568 to a single define_insn_and_split pattern.
3570 2018-10-05 David Malcolm <dmalcolm@redhat.com>
3573 * input.c (expand_location_to_spelling_point): Add param "aspect"
3574 and use rather than hardcoding LOCATION_ASPECT_CARET.
3575 (get_substring_ranges_for_loc): Handle the case of a single token
3576 within a macro expansion.
3577 * input.h (expand_location_to_spelling_point): Add "aspect" param,
3578 defaulting to LOCATION_ASPECT_CARET.
3580 2018-10-05 Paul Koning <ni1d@arrl.net>
3582 * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
3583 (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
3584 (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
3585 (pdp11_guard_type): New function.
3587 2018-10-05 Paul Koning <ni1d@arrl.net>
3589 * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
3590 * config/pdp11/pdp11.opt (mfloat32): Remove.
3592 * doc/invoke.texi (pdp11 -mfloat32): Remove:
3593 (pdp11 -mfloat64): Remove.
3595 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
3597 * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
3598 (*cmp<mode>_cc_i387): Ditto.
3599 (*cmpu<mode>_cc_i387): Ditto.
3600 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
3601 * config/i386/i386.c (ix86_expand_fp_compare): Remove
3603 <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
3604 Emit x86_sahf_1 pattern.
3605 (ix86_expand_compare): Update call to ix86_expand_fp_compare.
3606 (ix86_expand_carry_flag_compare): Ditto.
3608 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
3610 * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
3611 to reg_or_0_operand. Add "C" constraint.
3612 (*cmpxf_cc_i387): Ditto.
3613 (*cmp<mode>_i387): Change operand 2 predicate
3614 to nonimm_or_0_operand. Add "C" constraint.
3615 (*cmp<mode>_cc_i387): Ditto.
3616 (*cmp<mode>_0_i387): Remove insn pattern.
3617 (*cmp<mode>_0_cc_i387): Ditto.
3619 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
3621 * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
3622 * config/i386/predicates.md (nonimm_or_0_operand): Rename
3623 from vector_move_operand. Update all uses.
3625 2018-10-05 Martin Sebor <msebor@redhat.com>
3627 PR tree-optimization/87490
3628 * builtins.c (expand_builtin_strnlen): Handle a null data.decl
3631 2018-10-05 Richard Biener <rguenther@suse.de>
3633 PR tree-optimization/63155
3634 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
3635 vertical space in dumpfiles.
3636 * tree-ssa-propagate.h
3637 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
3638 * tree-ssa-propagate.c (cfg_blocks_back): New global.
3639 (ssa_edge_worklist_back): Likewise.
3640 (curr_order): Likewise.
3641 (cfg_blocks_get): Remove abstraction.
3642 (cfg_blocks_add): Likewise.
3643 (cfg_blocks_empty_p): Likewise.
3644 (add_ssa_edge): Add to current or next worklist based on
3646 (add_control_edge): Likewise.
3647 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
3649 (ssa_propagation_engine::ssa_propagate): ... here. Unify
3650 iteration from CFG and SSA edge worklist so we process
3651 everything in RPO order, prioritizing forward progress
3653 (ssa_prop_init): Allocate new worklists, do not dump
3655 (ssa_prop_fini): Free new worklists.
3657 2018-10-05 Richard Biener <rguenther@suse.de>
3659 * tree-core.h (tree_block::abstract_flag): Remove.
3660 (tree_block::block_num): Make full 32bits.
3661 * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
3662 * tree.h (BLOCK_ABSTRACT): Remove.
3663 * dwarf2out.c (gen_lexical_block_die): Remove dead code
3664 resulting from BLOCK_ABSTRACT being always false.
3665 (gen_inlined_subroutine_die): Likewise.
3666 (gen_block_die): Likewise.
3667 * tree.c (block_ultimate_origin): Likewise.
3668 * tree-pretty-print.c (dump_block_node): Remove code dealing
3669 with BLOCK_ABSTRACT.
3670 * tree-ssa-live.c (dump_scope_block): Likewise.
3671 * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
3672 * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
3674 2018-10-05 Richard Biener <rguenther@suse.de>
3676 * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
3677 is asked for initialize mode to the component mode of the
3680 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
3683 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
3684 assembler for -mavx.
3685 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
3687 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
3690 * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
3692 * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
3693 to be DImode. When using mffscrn, force the operand to a register.
3695 2018-10-04 Uros Bizjak <ubizjak@gmail.com>
3697 * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
3698 from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
3699 X87MODEF mode iterator.
3700 (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
3701 *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
3702 X87MODEF mode iterator.
3704 2018-10-04 Vinay Kumar <vinay.kumar@blackfigtech.com>
3706 * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
3707 -Wno-prio-ctor-dtor.
3709 2018-10-04 David Malcolm <dmalcolm@redhat.com>
3711 * Makefile.in (OBJS): Add opt-problem.o.
3712 * dump-context.h: Include "selftest.h.
3713 (selftest::temp_dump_context): New forward decl.
3714 (class dump_context): Make friend of class
3715 selftest::temp_dump_context.
3716 (dump_context::dump_loc_immediate): New decl.
3717 (class dump_pretty_printer): Move here from dumpfile.c.
3718 (class temp_dump_context): Move to namespace selftest.
3719 (temp_dump_context::temp_dump_context): Add param
3720 "forcibly_enable_dumping".
3721 (selftest::verify_dumped_text):
3722 (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
3723 (selftest::verify_item):
3724 (ASSERT_IS_TEXT): Move here from dumpfile.c.
3725 (ASSERT_IS_TREE): Likewise.
3726 (ASSERT_IS_GIMPLE): Likewise.
3727 * dumpfile.c (dump_context::dump_loc): Move immediate dumping
3729 (dump_context::dump_loc_immediate): ...this new function.
3730 (class dump_pretty_printer): Move to dump-context.h.
3731 (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
3732 (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
3733 (temp_dump_context::temp_dump_context): Move to "selftest"
3734 namespace. Add param "forcibly_enable_dumping", and use it to
3735 conditionalize the use of m_pp;
3736 (selftest::verify_dumped_text): Make non-static.
3737 (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
3738 (selftest::verify_item): Make non-static.
3739 (ASSERT_IS_TEXT): Move to dump-context.h.
3740 (ASSERT_IS_TREE): Likewise.
3741 (ASSERT_IS_GIMPLE): Likewise.
3742 (selftest::test_capture_of_dump_calls): Pass "true" for new
3743 param of temp_dump_context.
3744 * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
3745 it to MSG_ALL_PRIORITIES. Update values of TDF_COMPARE_DEBUG and
3747 * opt-problem.cc: New file.
3748 * opt-problem.h: New file.
3749 * optinfo-emit-json.cc
3750 (selftest::test_building_json_from_dump_calls): Pass "true" for
3751 new param of temp_dump_context.
3752 * optinfo.cc (optinfo_kind_to_dump_flag): New function.
3753 (optinfo::emit_for_opt_problem): New function.
3754 (optinfo::emit): Clarity which emit_item is used.
3755 * optinfo.h (optinfo::get_dump_location): New accessor.
3756 (optinfo::emit_for_opt_problem): New decl.
3757 (optinfo::emit): Make const.
3758 * selftest-run-tests.c (selftest::run_tests): Call
3759 selftest::opt_problem_cc_tests.
3760 * selftest.h (selftest::opt_problem_cc_tests): New decl.
3761 * tree-data-ref.c (dr_analyze_innermost): Convert return type from
3762 bool to opt_result, converting fprintf messages to
3763 opt_result::failure_at calls. Add "stmt" param for use by the
3765 (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
3766 (runtime_alias_check_p): Convert return type from bool to
3767 opt_result, converting dump_printf calls to
3768 opt_result::failure_at, using the statement DDR_A for their
3770 (find_data_references_in_stmt): Convert return type from bool to
3771 opt_result, converting "return false" to opt_result::failure_at
3773 * tree-data-ref.h: Include "opt-problem.h".
3774 (dr_analyze_innermost): Convert return type from bool to opt_result,
3775 and add a const gimple * param.
3776 (find_data_references_in_stmt): Convert return type from bool to
3778 (runtime_alias_check_p): Likewise.
3779 * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
3780 dr_analyze_innermost.
3781 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
3782 Convert return type from bool to opt_result, adding a message for
3783 the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
3784 (vect_analyze_data_ref_dependence): Convert return type from bool
3785 to opt_result. Change sense of return type from "false"
3786 effectively meaning "no problems" to "false" meaning a problem,
3787 so that "return false" becomes "return opt_result::success".
3788 Convert "return true" calls to opt_result::failure_at, using
3789 the location of statement A rather than vect_location.
3790 (vect_analyze_data_ref_dependences): Convert return type from bool
3792 (verify_data_ref_alignment): Likewise, converting dump_printf_loc
3793 calls to opt_result::failure_at, using the stmt location rather
3795 (vect_verify_datarefs_alignment): Convert return type from bool
3797 (vect_enhance_data_refs_alignment): Likewise. Split local "stat"
3798 into multiple more-tightly-scoped copies.
3799 (vect_analyze_data_refs_alignment): Convert return type from bool
3801 (vect_analyze_data_ref_accesses): Likewise, converting a
3802 "return false" to a "return opt_result::failure_at", adding a
3804 (vect_prune_runtime_alias_test_list): Convert return type from
3805 bool to opt_result, converting dump_printf_loc to
3806 opt_result::failure_at. Add a %G to show the pertinent statement,
3807 and use the stmt's location rather than vect_location.
3808 (vect_find_stmt_data_reference): Convert return type from
3809 bool to opt_result, converting dump_printf_loc to
3810 opt_result::failure_at, using stmt's location.
3811 (vect_analyze_data_refs): Convert return type from bool to
3812 opt_result. Convert "return false" to "return
3813 opt_result::failure_at", adding messages as needed.
3814 * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
3815 type from bool to opt_result.
3816 (vect_determine_vf_for_stmt): Likewise.
3817 (vect_determine_vectorization_factor): Likewise, converting
3818 dump_printf_loc to opt_result::failure_at, using location of phi
3819 rather than vect_location.
3820 (vect_analyze_loop_form_1): Convert return type from bool to
3821 opt_result, converting dump_printf_loc calls, retaining the use of
3823 (vect_analyze_loop_form): Convert return type from loop_vec_info
3824 to opt_loop_vec_info.
3825 (vect_analyze_loop_operations): Convert return type from bool to
3826 opt_result, converting dump_printf_loc calls, using the location
3827 of phi/stmt rather than vect_location where available. Convert
3828 various "return false" to "return opt_result::failure_at" with
3829 "unsupported phi" messages.
3830 (vect_get_datarefs_in_loop): Convert return type from bool to
3831 opt_result. Add a message for the
3832 PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
3833 (vect_analyze_loop_2): Convert return type from bool to
3834 opt_result. Ensure "ok" is set to a opt_result::failure_at before
3835 each "goto again;", adding new messages where needed.
3836 Add "unsupported grouped {store|load}" messages.
3837 (vect_analyze_loop): Convert return type from loop_vec_info to
3839 * tree-vect-slp.c (vect_analyze_slp): Convert return type from
3841 * tree-vect-stmts.c (process_use): Likewise, converting
3842 dump_printf_loc call and using stmt location, rather than
3844 (vect_mark_stmts_to_be_vectorized): Likeise.
3845 (vect_analyze_stmt): Likewise, adding a %G.
3846 (vect_get_vector_types_for_stmt): Convert return type from bool to
3847 opt_result, converting dump_printf_loc calls and using stmt
3848 location, rather than vect_location.
3849 (vect_get_mask_type_for_stmt): Convert return type from tree to
3850 opt_tree, converting dump_printf_loc calls and using stmt location.
3851 * tree-vectorizer.c: Include "opt-problem.h.
3852 (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
3853 MSG_PRIORITY_INTERNALS. Convert local "loop_vinfo" from
3854 loop_vec_info to opt_loop_vec_info. If if fails, and dumping is
3855 enabled, use it to report at the top level "couldn't vectorize
3856 loop" followed by the problem.
3857 * tree-vectorizer.h (opt_loop_vec_info): New typedef.
3858 (vect_mark_stmts_to_be_vectorized): Convert return type from bool
3860 (vect_analyze_stmt): Likewise.
3861 (vect_get_vector_types_for_stmt): Likewise.
3862 (tree vect_get_mask_type_for_stmt): Likewise.
3863 (vect_analyze_data_ref_dependences): Likewise.
3864 (vect_enhance_data_refs_alignment): Likewise.
3865 (vect_analyze_data_refs_alignment): Likewise.
3866 (vect_verify_datarefs_alignment): Likewise.
3867 (vect_analyze_data_ref_accesses): Likewise.
3868 (vect_prune_runtime_alias_test_list): Likewise.
3869 (vect_find_stmt_data_reference): Likewise.
3870 (vect_analyze_data_refs): Likewise.
3871 (vect_analyze_loop): Convert return type from loop_vec_info to
3873 (vect_analyze_loop_form): Likewise.
3874 (vect_analyze_slp): Convert return type from bool to opt_result.
3876 2018-10-04 David Malcolm <dmalcolm@redhat.com>
3878 * doc/invoke.texi (-fopt-info): Document new "internals"
3880 * dump-context.h (dump_context::apply_dump_filter_p): New decl.
3881 * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
3883 (optinfo_verbosity_options): Add "internals".
3884 (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
3885 (dump_context::apply_dump_filter_p): New member function.
3886 (dump_context::dump_loc): Use apply_dump_filter_p rather than
3887 explicitly masking the dump_kind.
3888 (dump_context::begin_scope): Increment the scope depth first. Use
3889 apply_dump_filter_p rather than explicitly masking the dump_kind.
3890 (dump_context::emit_item): Use apply_dump_filter_p rather than
3891 explicitly masking the dump_kind.
3892 (dump_dec): Likewise.
3893 (dump_hex): Likewise.
3894 (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
3895 (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
3896 (opt_info_switch_p): Update handling of default
3897 MSG_OPTIMIZED_LOCATIONS to cope with default of
3898 MSG_PRIORITY_USER_FACING.
3899 (dump_basic_block): Use apply_dump_filter_p rather than explicitly
3900 masking the dump_kind.
3901 (selftest::test_capture_of_dump_calls): Update test_dump_context
3902 instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
3903 than MSG_ALL. Generalize scope test to be run at all four
3904 combinations of with/without MSG_PRIORITY_USER_FACING and
3905 MSG_PRIORITY_INTERNALS, adding examples of explicit priority
3906 for each of the two values.
3907 * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
3908 Rename MSG_ALL to MSG_ALL_KINDS. Add MSG_PRIORITY_USER_FACING,
3909 MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
3910 values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
3911 (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
3912 with MSG_PRIORITY_*.
3913 * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
3914 dump messages as MSG_PRIORITY_USER_FACING.
3915 * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
3916 about the interaction with MSG_PRIORITY_*.
3918 2018-10-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
3920 * varasm.c (output_constant): Add new parameter merge_strings.
3921 Make strings properly zero terminated in merge string sections.
3922 (mergeable_string_section): Don't fail if the last char is non-zero.
3923 (assemble_variable_contents): Handle merge string sections.
3924 (assemble_variable): Likewise.
3925 (assemble_constant_contents): Likewise.
3926 (output_constant_def_contents): Likewise.
3927 (output_constructor_array_range,
3928 output_constructor_regular_field): Adjust call to output_constant.
3929 (output_object_block): Adjust call to assemble_constant_contents
3930 and assemble_variable_contents.
3932 2018-10-04 Martin Liska <mliska@suse.cz>
3935 * cgraphunit.c (process_function_and_variable_attributes):
3936 Warn about a function with alias attribute and a body.
3938 2018-10-04 Martin Liska <mliska@suse.cz>
3941 * multiple_target.c (redirect_to_specific_clone): New function.
3942 (ipa_target_clone): Use it.
3943 * tree-inline.c: Fix comment.
3945 2018-10-04 David Malcolm <dmalcolm@redhat.com>
3947 * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
3949 (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
3950 (gcc::dump_manager::register_pass): New member function, adapted
3951 from loop body in gcc::pass_manager::register_pass, adding a
3952 call to update_dfi_for_opt_info.
3953 (gcc::dump_manager::opt_info_enable_passes): Store the
3954 -fopt-info options into the new fields. Move the loop
3956 (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
3958 * dumpfile.h (struct opt_pass): New forward decl.
3959 (gcc::dump_manager::register_pass): New decl.
3960 (gcc::dump_manager::update_dfi_for_opt_info): New decl.
3961 (class gcc::dump_manager): Add fields "m_optgroup_flags",
3962 "m_optinfo_flags", and "m_optinfo_filename".
3963 * passes.c (gcc::pass_manager::register_pass): Move all of the
3964 dump-handling code to gcc::dump_manager::register_pass.
3966 2018-10-04 Peter Bergner <bergner@linux.ibm.com>
3968 PR rtl-optimization/87466
3969 * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
3970 * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
3971 * doc/tm.texi: Regenerate.
3972 * ira-lives.c (process_bb_node_lives): Use the new target hook.
3973 * lra-lives.c (process_bb_lives): Likewise.
3974 * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
3977 2018-10-04 Tamar Christina <tamar.christina@arm.com>
3979 * params.c (add_params): Fix initialization.
3981 2018-10-04 Martin Liska <mliska@suse.cz>
3983 PR gcov-profile/84107
3984 * tree-profile.c (init_ic_make_global_vars):
3985 Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
3986 Come up with new ic_tuple* variables. Emit
3987 __gcov_indirect_call{,_topn} variables.
3988 (gimple_gen_ic_profiler): Access the variable
3990 (gimple_gen_ic_func_profiler): Access
3991 __gcov_indirect_call.callee field.
3992 (gimple_init_gcov_profiler): Use ptr_type_node.
3993 * value-prof.c (gimple_ic): Use ptr_type_node.
3995 2018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3997 PR tree-optimization/85787
3998 * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
3999 into this function and add support for detecting multiple phis.
4000 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
4002 2018-10-04 Martin Liska <mliska@suse.cz>
4005 * ipa-inline.c (inline_to_all_callers_1):
4006 Call ultimate_alias_target for node being inlined.
4008 2018-10-03 Jeff Law <law@redhat.com>
4010 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
4012 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
4013 * tree.h (get_typenode_from_name): Prototype.
4015 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
4017 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
4018 Change operand 2 predicate to nonimmediate_operand.
4019 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4021 2018-10-03 Martin Sebor <msebor@redhat.com>
4022 Jeff Law <law@redhat.com>
4024 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
4026 (get_string_length): Detect unterminated arrays.
4027 (format_string): Same.
4028 (format_directive): Warn about unterminated arrays.
4029 (handle_gimple_call): Mark statements with no_warning as needed.
4031 2018-10-03 Jim Wilson <jimw@sifive.com>
4033 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
4034 also define __riscv_abi_rve. Delete trailing white space.
4036 2018-10-03 Paul Koning <ni1d@arrl.net>
4038 Enable LRA register allocator for PDP11.
4039 * config/pdp11/constraints.md (Q): Use define_memory_constraint.
4042 * config/pdp11/pdp11.c (pdp11_lra_p): New function.
4043 * config/pdp11/pdp11.opt (-mlra): New option.
4044 * doc/invoke.texi (PDP-11 Options): Document -mlra.
4046 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
4048 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
4049 (*<absneg:code>extend<mode>xf2): Ditto.
4051 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
4053 PR tree-optimization/87415
4054 * tree-vrp.c (set_value_range_with_overflow): Special case one bit
4057 2018-10-02 Jeff Law <law@redhat.com>
4059 * gimple-fold.c (get_range_strlen): Only set *nonstr when
4060 an unterminated string is discovered. Bubble up range
4061 even for unterminated strings.
4062 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
4063 indicates the string was not terminated via NONSTR.
4065 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
4067 * tree-vrp.c (extract_range_from_unary_expr): Special case all
4068 pointer conversions.
4069 Do not do anything special for anti-ranges.
4071 2018-10-03 Jérôme Lambourg <lambourg@adacore.com>
4073 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
4074 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
4076 2018-10-03 Martin Liska <mliska@suse.cz>
4078 PR gcov-profile/86109
4079 * coverage.c (coverage_begin_function): Do not
4080 mark lambdas as artificial.
4081 * tree-core.h (struct GTY): Remove tm_clone_flag
4082 and introduce new lambda_function.
4083 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
4085 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
4088 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
4089 P8_VECTOR and VSX are enabled.
4091 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
4093 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
4094 0x3907 as CPU model number.
4096 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
4098 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
4099 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
4100 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
4101 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
4102 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
4103 * config/s390/s390.md: Likewise. Rename also the cpu attribute
4104 value from arch12 to z14.
4106 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
4108 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
4110 (isinf<mode>2): Ditto.
4112 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
4114 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
4115 before emitting fxam. Perform calculations in XFmode.
4117 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
4119 * match.pd (((X /[ex] A) +- B) * A): New transformation.
4121 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
4124 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
4125 * tree.c (signed_or_unsigned_type_for): Handle complex.
4127 2018-10-02 Jeff Law <law@redhat.com>
4129 * gimple-fold.c (get_range_strlen): Remove dead code.
4131 2018-10-02 Martin Sebor <msebor@redhat.com>
4132 Jeff Law <law@redhat.com>
4134 * builtins.c (unterminated_array): Add new arguments.
4135 If argument is not terminated, bubble up size and exact
4137 (expand_builtin_strnlen): Detect, avoid expanding
4138 and diagnose unterminated arrays.
4139 (c_strlen): Fill in offset of start of unterminated strings.
4140 * builtins.h (unterminated_array): Update prototype.
4142 2018-10-02 Richard Biener <rguenther@suse.de>
4144 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
4145 of haddv4df, first reduce to SSE width and exploit the fact
4146 that we only need element zero with the reduction result.
4147 (reduc_plus_scal_v2df): Likewise.
4149 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
4151 * dojump.h (do_jump): Delete.
4152 (do_jump_1): Likewise.
4153 (split_comparison): Move around.
4154 * dojump.c (do_jump): Make static.
4155 (do_jump_1): Likewise.
4156 (jumpifnot): Move around.
4157 (jumpifnot_1): Likewise.
4159 (jumpif_1): Likewise.
4160 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
4162 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
4164 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
4165 insns in the delay slot and add_insn_after for the jump insn.
4167 2018-10-02 Richard Biener <rguenther@suse.de>
4169 * tree-inline.c (expand_call_inline): Use the location of
4170 the callee declaration for the inline-entry marker.
4171 * final.c (notice_source_line): Remove special-casing of
4172 NOTE_INSN_INLINE_ENTRY.
4174 2018-10-01 Carl Love <cel@us.ibm.com>
4177 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
4178 (__builtin_mtfsb0): New.
4179 (__builtin_mtfsb1): New.
4180 ( __builtin_set_fpscr_rn): New.
4181 (__builtin_set_fpscr_drn): New.
4182 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
4183 (rs6000_expand_set_fpscr_rn_builtin): Add.
4184 (rs6000_expand_set_fpscr_drn_builtin): Add.
4185 (rs6000_expand_builtin): Add case statement entries for
4186 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
4187 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
4188 RS6000_BUILTIN_MFFSL.
4189 (rs6000_init_builtins): Add ftype initialization and def_builtin
4190 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
4191 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
4192 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
4193 rs6000_mffscdrn): Add define_insn.
4194 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
4195 * doc/extend.texi: Add documentation for the builtins.
4197 2018-10-01 Richard Biener <rguenther@suse.de>
4199 PR tree-optimization/87465
4200 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
4201 causing branch miscounts.
4203 2018-10-01 Tamar Christina <tamar.christina@arm.com>
4205 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
4206 aarch64_option_default_param): New.
4207 (params.h): Include.
4208 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
4209 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
4210 stack-clash protection validation code.
4212 2018-10-01 Tamar Christina <tamar.christina@arm.com>
4214 * params.c (validate_param): New.
4215 (add_params): Use it.
4216 (set_param_value): Refactor param validation into validate_param.
4217 (diagnostic.h): Include.
4218 * diagnostic.h (diagnostic_ready_p): New.
4220 2018-10-01 Tamar Christina <tamar.christina@arm.com>
4222 * params.c (set_param_value):
4223 Add index of parameter being validated.
4224 * common/common-target.def (option_validate_param): New.
4225 * common/common-targhooks.h (default_option_validate_param): New.
4226 * common/common-targhooks.c (default_option_validate_param): New.
4227 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
4228 * doc/tm.texi: Regenerate.
4230 2018-10-01 Tamar Christina <tamar.christina@arm.com>
4233 * config/aarch64/aarch64.c (aarch64_override_options_internal):
4234 Add validation for stack-clash parameters and set defaults.
4236 2018-10-01 Tamar Christina <tamar.christina@arm.com>
4239 * configure.ac: Add stack-clash-protection-guard-size.
4240 * doc/install.texi: Document it.
4241 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
4242 * params.def: Update comment for guard-size.
4243 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
4244 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
4245 * configure: Regenerate.
4247 2018-10-01 Tamar Christina <tamar.christina@arm.com>
4250 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
4251 STACK_DYNAMIC_OFFSET): New.
4252 * config/aarch64/aarch64.c (aarch64_layout_frame):
4253 Update outgoing args size.
4254 (aarch64_stack_clash_protection_alloca_probe_range,
4255 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
4257 2018-10-01 Tamar Christina <tamar.christina@arm.com>
4260 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
4262 * target.def (stack_clash_protection_alloca_probe_range): New.
4263 (stack_clash_protection_final_dynamic_probe): Remove.
4264 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
4265 (default_stack_clash_protection_final_dynamic_probe): Remove.
4266 * targhooks.c: Likewise.
4267 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
4268 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
4269 * doc/tm.texi: Regenerate.
4271 2018-10-01 Tamar Christina <tamar.christina@arm.com>
4274 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
4275 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
4276 aarch64_clamp_to_uimm12_shift): New.
4277 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
4278 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
4280 2018-10-01 Tamar Christina <tamar.christina@arm.com>
4283 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
4285 2018-10-01 Jeff Law <law@redhat.com>
4286 Richard Sandiford <richard.sandiford@linaro.org>
4287 Tamar Christina <tamar.christina@arm.com>
4290 * config/aarch64/aarch64.md
4291 (probe_stack_range): Add k (SP) constraint.
4292 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
4293 STACK_CLASH_MAX_UNROLL_PAGES): New.
4294 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
4295 stack probes for stack clash.
4296 (aarch64_allocate_and_probe_stack_space): New.
4297 (aarch64_expand_prologue): Use it.
4298 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
4299 (aarch64_sub_sp): Add emit_move_imm optional param.
4301 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
4303 PR tree-optimization/87261
4304 * match.pd: Remove trailing whitespace.
4305 Add (x & y) | ~(x | y) -> ~(x ^ y),
4306 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
4308 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
4310 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
4312 (ashlsi3_insn): Update instruction constraints.
4313 (ashrsi3_insn): Likewise.
4314 (rotrsi3): Likewise.
4315 (add_shift): Likewise.
4316 * config/arc/constraints.md (Csz): New 32 bit constraint. It
4317 avoids placing in the limm field small constants which, otherwise,
4318 could end into a small instruction.
4320 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
4322 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
4323 destination register is not odd-even.
4324 (umaddsidi4_split): Likewise.
4326 2018-10-01 Richard Biener <rguenther@suse.de>
4328 * tree-inline.c (expand_call_inline): Store origin of fn
4329 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
4330 * tree.c (block_ultimate_origin): Simplify and do some
4333 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
4335 * config/i386/mmx.md (EMMS): New int iterator.
4336 (emms): New int attribute.
4337 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
4338 EMMS int iterator. Explicitly declare clobbers.
4339 (mmx_emms): Remove expander.
4341 * config/i386/predicates.md (emms_operation): Remove predicate.
4342 (vzeroall_pattern): New predicate.
4343 (vzeroupper_pattern): Rename from vzeroupper_operation.
4344 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
4345 vzeroupper_pattern and vzeroall_pattern predicates.
4347 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
4349 PR rtl-optimization/86939
4350 * ira-lives.c (make_hard_regno_born): Rename from this...
4351 (make_hard_regno_live): ... to this. Remove update to conflict
4352 information. Update function comment.
4353 (make_hard_regno_dead): Add conflict information update. Update
4355 (make_object_born): Rename from this...
4356 (make_object_live): ... to this. Remove update to conflict information.
4357 Update function comment.
4358 (make_object_dead): Add conflict information update. Update function
4360 (mark_pseudo_regno_live): Call make_object_live.
4361 (mark_pseudo_regno_subword_live): Likewise.
4362 (mark_hard_reg_dead): Update function comment.
4363 (mark_hard_reg_live): Call make_hard_regno_live.
4364 (process_bb_node_lives): Likewise.
4365 * lra-lives.c (make_hard_regno_born): Rename from this...
4366 (make_hard_regno_live): ... to this. Remove update to conflict
4367 information. Remove now uneeded check_pic_pseudo_p argument.
4368 Update function comment.
4369 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
4370 to conflict information. Update function comment.
4371 (mark_pseudo_live): Remove update to conflict information. Update
4373 (mark_pseudo_dead): Add conflict information update.
4374 (mark_regno_live): Call make_hard_regno_live.
4375 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
4376 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
4378 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
4381 * config/i386/i386.c (construct_container): Use TImode for
4382 BLKmode values in 2 integer registers.
4384 2018-09-29 Jeff Law <law@redhat.com>
4386 * builtins.c (unterminated_array): Pass in c_strlen_data * to
4387 c_strlen rather than just a tree *.
4388 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
4389 Update recursive calls appropriately. If caller did not provide a
4390 suitable data pointer, create a local one. When a non-terminated
4391 string is discovered, bubble up information about the string via the
4392 c_strlen_data object.
4393 * builtins.h (c_strlen): Update prototype.
4394 (c_strlen_data): New structure.
4395 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
4396 For a type 2 call, if c_strlen indicates a non-terminated string
4397 use the length of the non-terminated string.
4398 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
4400 2018-09-29 Jakub Jelinek <jakub@redhat.com>
4403 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
4404 __m512d type for __A argument rather than __m512.
4406 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
4408 * match.pd (simple_comparison): Don't optimize if either operand is
4409 a function pointer when target needs function pointer canonicalization.
4411 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
4413 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
4414 power5 .. power9 to remove indirection.
4415 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
4416 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
4417 ASM_CPU_476_SPEC): Delete.
4418 (ASM_CPU_SPEC): Adjust.
4419 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
4420 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
4422 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
4424 * config.in: Delete HAVE_AS_DCI.
4425 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
4426 * config/rs6000/rs6000.h: Ditto.
4427 * configure.ac: Delete HAVE_AS_DCI.
4428 * configure: Regenerate.
4430 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
4432 * config.in (HAVE_AS_LWSYNC): Delete.
4433 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
4434 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
4436 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
4437 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
4439 * configure.ac: Delete HAVE_AS_LWSYNC.
4440 * configure: Regenerate.
4442 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
4443 Pierre-Marie de Rodat <derodat@adacore.com>
4445 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
4446 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
4447 (cgraph_node::create_thunk): Add indirect_offset parameter.
4448 (thunk_adjust): Likewise.
4449 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
4450 and initialize the corresponding field with it.
4451 (cgraph_node::dump): Dump indirect_offset field.
4452 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
4453 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
4454 (thunk_adjust): Add indirect_offset parameter and deal with it.
4455 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
4456 pass it to thunk_adjust. Do not call the target hook if it's non-zero
4457 or if the thunk is external or local. Fix formatting. Do not chain
4458 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
4459 if any, in the GIMPLE representation.
4460 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
4461 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
4462 (input_node): Read indirect_offset field.
4463 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
4464 call to thunk_adjust.
4465 * tree-nested.c (struct nesting_info): Add thunk_p field.
4466 (create_nesting_tree): Set it.
4467 (convert_all_function_calls): Copy static chain from targets to thunks.
4468 (finalize_nesting_tree_1): Return early for thunks.
4469 (unnest_nesting_tree_1): Do not finalize thunks.
4470 (gimplify_all_functions): Do not gimplify thunks.
4472 2018-09-28 David Malcolm <dmalcolm@redhat.com>
4474 * opt-suggestions.c (option_proposer::build_option_suggestions):
4475 Release "option_values".
4477 2018-09-28 David Malcolm <dmalcolm@redhat.com>
4479 * coverage.c (get_coverage_counts): Convert problem-reporting dump
4480 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
4481 * dumpfile.c (kind_as_string): New function.
4482 (dump_loc): Rather than a hardcoded prefix of "note: ", use
4483 kind_as_string to vary the prefix based on dump_kind.
4484 (selftest::test_capture_of_dump_calls): Update for above.
4486 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
4488 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
4489 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
4491 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
4493 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
4494 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
4495 INVALID_REGNUM instead of FPSR_REG.
4496 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
4497 * config/i386/i386.md: Update comment of FP compares.
4498 (fldenv): Do not clobber FPSR_REG.
4500 2018-09-28 Richard Biener <rguenther@suse.de>
4502 * tree.h (BLOCK_ORIGIN): New.
4503 * omp-expand.c (grid_expand_target_grid_body): Assign
4504 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
4505 * tree-inline.c (remap_block): Likewise.
4506 * auto-profile.c (get_function_decl_from_block): Simplify
4507 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
4508 * langhooks.c (lhd_print_error_function): Likewise.
4509 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
4511 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4512 * tree.c (block_nonartificial_location): Likewise.
4513 (block_ultimate_origin): Likewise.
4514 * tree-pretty-print.c (percent_K_format): Likewise. Remove
4515 no longer needed LTO case.
4517 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
4518 Jan Hubicka <jh@suse.cz>
4519 Martin Jambor <mjambor@suse.cz>
4521 * simplify-rtx.c (simplify_merge_mask): New function.
4522 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
4523 same masks are used in op1 or op2.
4524 (test_vec_merge): New function.
4525 (test_vector_ops): Call test_vec_merge.
4527 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
4529 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
4530 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
4531 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
4532 (cypress_costs): Set it.
4533 (supersparc_costs): Likewise.
4534 (hypersparc_costs): Likewise.
4535 (leon_cost): Likewise.
4536 (leon3_costs): Likewise.
4537 (sparclet_costs): Likewise.
4538 (ultrasparc_costs): Likewise.
4539 (ultrasparc_costs): Likewise.
4540 (niagara_costs): Likewise.
4541 (niagara2_costs): Likewise.
4542 (niagara3_costs): Likewise.
4543 (niagara4_costs): Likewise.
4544 (niagara7_costs): Likewise.
4545 (m8_costs): Likewise.
4546 (TARGET_CAN_FOLLOW_JUMP): Define.
4547 (pass_work_around_errata::gate): Minor tweak.
4548 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
4549 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
4550 Automaitcally clear MASK_FSMULD mask for V7 processors.
4551 (sparc_can_follow_jump): New static function.
4552 (output_ubranch): Deal with CROSSING_JUMP_P.
4553 (sparc_use_sched_lookahead): Rewrite using switch statement.
4554 (sparc_issue_rate): Reorder.
4555 (sparc_branch_cost): New function.
4557 2018-09-27 Martin Sebor <msebor@redhat.com>
4559 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
4560 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
4561 (int_fits_type_p): Same.
4563 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
4565 * config/i386/i386.md (FPCR_REG): Remove.
4566 (UNSPEC_FLDCW): Remove.
4567 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
4568 (x86_fldcw_1): Remove insn pattern.
4569 (fnstenv): Do not clobber FPCR_REG.
4571 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
4572 (CALL_USED_REGISTERS): Ditto.
4573 (REG_ALLOC_ORDER): Ditto.
4574 (REG_CLASS_CONTENTS): Ditto.
4575 (HI_REGISTER_NAMES): Ditto.
4576 (ADDITIONAL_REGISTER_NAMES): Use defines instead
4577 of numerical constants.
4578 * config/i386/i386.c (regclass_map): Remove fpsr register.
4579 (dbx_register_map): Ditto.
4580 (dbx64_register_map): Ditto.
4581 (svr4_dbx_register_map): Ditto.
4582 (print_reg): Do not handle FPCR_REG.
4584 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
4587 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
4588 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
4589 Delete, always treat as true.
4590 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
4591 Ditto. Simplify remaining code.
4592 * config/powerpcspe/powerpcspe.h: Ditto.
4593 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
4594 Simplify remaining code.
4595 (rs6000_expand_builtin): Ditto.
4596 * config/rs6000/rs6000.h: Ditto.
4597 * configure.ac: Ditto.
4598 * configure: Regenerate.
4600 2018-09-27 Martin Liska <mliska@suse.cz>
4602 * coverage.c (get_coverage_counts): Revert the formatting
4603 of missing profile opt info.
4605 2018-09-27 Richard Biener <rguenther@suse.de>
4609 * dwarf2out.c (set_block_origin_self): Do not mark outermost
4610 block as we do not output that.
4611 (gen_inlined_subroutine_die): Elide the originally outermost
4612 block, matching what we do for concrete instances.
4613 (decls_for_scope): Add parameter specifying whether to recurse
4616 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
4617 Tom de Vries <tom@codesourcery.com>
4621 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
4622 STORE_FLAG_VALUE == 1.
4624 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
4626 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
4627 constant definitions.
4628 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
4629 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
4630 ("speculation_barrier"): New expander definition.
4632 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
4634 PR gcov-profile/86957
4635 * common.opt: New warning option -Wmissing-profile.
4636 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
4637 * doc/invoke.texi: Document -Wmissing-profile.
4639 2018-09-26 Jim Wilson <jimw@sifive.com>
4641 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
4642 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
4644 2018-09-26 Martin Sebor <msebor@redhat.com>
4646 * tree.c (zerop): Change return type to bool.
4647 (integer_zerop, integer_onep, integer_each_onep): Same.
4648 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
4649 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
4650 (real_onep, real_minus_onep, chain_index): Same.
4651 (print_type_hash_statistics, type_list_equal): Same.
4652 * tree.h (zerop): Same.
4653 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
4654 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
4655 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
4656 (real_onep, real_minus_onep, chain_index): Same.
4657 (print_type_hash_statistics, type_list_equal): Same.
4659 2018-09-26 Jim Wilson <jimw@sifive.com>
4661 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
4663 2018-09-26 Jakub Jelinek <jakub@redhat.com>
4666 * config/i386/i386.c: Include debug.h and dwarf2out.h.
4667 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
4670 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
4672 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
4674 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
4676 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
4677 and FP_SECOND_SSE_REGS.
4678 (REG_CLASS_NAMES): Ditto.
4679 (REG_CLASS_CONTENTS): Ditto.
4680 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
4681 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
4682 (ix86_preferred_output_reload_class): Ditto.
4683 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
4684 clobber constraint to "=&f".
4685 (fix_truncdi_i387): Ditto.
4686 (lrintxfdi2): Ditto.
4687 (fistdi2_<rounding>): Ditto.
4688 (fpremxf4_i387): Change "=u" constraint to "=f".
4689 (fprem1xf4_i387): Ditto.
4691 (fptanxf4_i387): Ditto.
4692 (fxtractxf3_i387): Ditto.
4693 (fscalexf4_i387): Ditto.
4694 (atan2xf3): Change "u" constraint to "f".
4695 (fyl2xxf3_i387): Ditto.
4696 (fyl2xp1xf3_i387): Ditto.
4698 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
4701 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
4702 for removed I387_MASK_PM entity.
4705 2018-09-26 Jeff Law <law@redhat.com>
4707 2018-09-26 Alexey Neyman <stilor@att.net>
4709 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
4710 headers are no longer pulled in by <isl/val.h>.
4712 2018-09-26 Richard Biener <rguenther@suse.de>
4715 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
4716 or concrete instance DIE to the tree. Create abstract origin
4717 attributes also for concrete instances.
4719 2018-09-26 Alexey Neyman <stilor@att.net>
4721 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
4722 headers are no longer pulled in by <isl/val.h>.
4724 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
4726 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
4727 Use new helper functions.
4728 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
4729 Use new helper functions.
4730 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
4731 aarch_mm_needs_release): New declarations.
4732 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
4733 aarch_mm_needs_release): New.
4735 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
4737 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
4738 (arm32_output_mi_thunk): Deal with long calls.
4740 2018-09-26 Richard Biener <rguenther@suse.de>
4744 * tree-inline.c (expand_call_inline): When the location
4745 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
4746 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
4747 the inserted BLOCK to make inlined_function_outer_scope_p
4749 * dwarf2out.c (add_call_src_coords_attributes): Do not add
4750 coords for reserved locations.
4752 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
4754 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
4755 (*call_indirect_nonlocal_sysv<mode>): Ditto.
4756 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
4757 (*sibcall_nonlocal_sysv<mode>): Ditto.
4758 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
4759 (<bd>_<mode>): Ditto.
4760 (<bd>tf_<mode>): Ditto.
4762 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
4764 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
4765 control string as a list of templates instead of as C code.
4766 (*altivec_movti): Ditto.
4767 * config/rs6000/darwin.md (movdf_low_di): Ditto.
4769 2018-09-25 Jim Wilson <jimw@sifive.com>
4771 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
4772 when target symbol is weak.
4774 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
4777 * builtins.c (unterminated_array): Simplify.
4778 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
4779 where pointer arithmetic is safe.
4781 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
4784 * config/rs6000/altivec.md (altivec_vspltb): Use
4785 const_0_to_15_operand instead of u5bit_cint_operand.
4786 (*altivec_vspltb_internal): Ditto.
4787 (altivec_vspltb_direct): Ditto.
4788 (altivec_vsplth): Use const_0_to_7_operand instead of
4790 (*altivec_vsplth_internal): Ditto.
4791 (altivec_vsplth_direct): Ditto.
4792 (altivec_vspltw): Use const_0_to_3_operand instead of
4794 (*altivec_vspltw_internal): Ditto.
4795 (altivec_vspltw_direct): Ditto.
4796 (altivec_vspltsf): Ditto.
4797 (*altivec_vspltsf_internal): Ditto.
4798 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
4799 various splats with the proper size immediate. Reorder the various
4800 cases by ascending size of immediate, and put all such together.
4802 2018-09-25 Richard Biener <rguenther@suse.de>
4805 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
4807 (maybe_create_die_with_external_ref): Do not create
4808 DW_TAG_imported_unit here.
4809 (add_abstract_origin_attribute): Handle external BLOCK refs.
4810 (dwarf2out_abstract_function): Simplify LTO case.
4811 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
4812 rather than using maybe_create_die_with_external_ref.
4814 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
4817 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
4818 (frndintxf2_mask_pm_i387): Ditto.
4819 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
4820 Enable for !flag_trapping_math.
4821 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
4822 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
4823 Change operand 1 predicate to nonimmediate_operand.
4824 (attr "i387_cw"): Remove mask_pm.
4825 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
4826 (enum ix86_entity): Remove I387_MASK_PM.
4827 * config/i386/i386.c (ix86_i387_mode_needed): Do not
4828 handle I387_MASK_PM.
4829 (ix86_mode_needed): Ditto.
4830 (ix86_mode_after): Ditto.
4831 (ix86_mode_entry): Ditto.
4832 (ix86_mode_exit): Ditto.
4833 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
4835 2018-09-25 Jakub Jelinek <jakub@redhat.com>
4837 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
4838 to_update_switch_stmts to vNULL instead of calling create on them
4841 2018-09-25 Richard Biener <rguenther@suse.de>
4843 PR tree-optimization/87402
4844 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
4845 (visit_phi): Re-instantiate handling of supposed to be VARYING
4846 but non-VARYING backedge value.
4848 2018-09-25 Richard Biener <rguenther@suse.de>
4851 * dwarf2out.c (struct sym_off_pair): New.
4852 (external_die_map): New global.
4853 (lookup_decl_die): When in LTO create DIEs lazily from the
4855 (lookup_block_die): New function, create DIEs lazily in LTO.
4856 (equate_block_to_die): New function.
4857 (dwarf2out_die_ref_for_decl): During WPA get the association
4858 from the external DIE map.
4859 (dwarf2out_register_external_die): Record mapping into the
4861 (maybe_create_die_with_external_ref): New function split out from
4862 DIE generation part of old dwarf2out_register_external_die.
4863 (add_abstract_origin_attribute): Do not return the DIE. When
4864 in LTO reference externals directly.
4865 (dwarf2out_abstract_function): When in LTO ignore calls for
4866 decls with external DIEs (already present abstract instances).
4867 (gen_call_site_die): Adjust.
4868 (add_high_low_attributes): Likewise.
4869 (gen_lexical_block_die): Likewise.
4870 (gen_inlined_subroutine_die): Likewie.
4871 (gen_block_die): Likewise.
4872 (dwarf2out_inline_entry): Likewise.
4873 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
4876 2018-09-25 Martin Liska <mliska@suse.cz>
4878 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
4879 integers and not by a float value.
4881 2018-09-25 Martin Liska <mliska@suse.cz>
4884 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
4886 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
4887 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
4888 * godump.c (go_format_type): Remove extra parenthesis.
4890 2018-09-25 Martin Liska <mliska@suse.cz>
4892 * alias.c (set_dest_equal_p): Remove unused function.
4893 * config/i386/i386.c (def_builtin_pure2): Likewise.
4894 * diagnostic-show-locus.c (class layout): Remove
4896 (layout::layout): Likewise here.
4897 * dump-context.h (class temp_dump_context): Likewise.
4898 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
4899 (add_AT_loclistsptr): Likewise.
4900 (add_AT_offset): Likewise.
4901 (get_AT_hi_pc): Likewise.
4902 (is_comdat_die): Likewise.
4903 (type_is_enum): Likewise.
4904 (ceiling): Likewise.
4905 (add_AT_vms_delta): Likewise.
4906 (is_class_die): Likewise.
4907 * edit-context.c (class line_event): Remove unused field.
4908 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
4910 * ipa-cp.c (ipa_get_vr_lat): Likewise.
4911 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
4912 (ok_for_base_p_nonstrict): Likewise.
4913 * tree-chrec.c (is_not_constant_evolution): Likewise.
4914 (chrec_fold_poly_cst): Likewise.
4915 * tree-if-conv.c (has_pred_critical_p): Likewise.
4916 * tree-ssa-coalesce.c (print_exprs): Likewise.
4917 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
4918 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
4919 * tree-vrp.c (value_ranges_intersect_p): Likewise.
4920 (value_range_nonnegative_p): Likewise.
4922 2018-09-25 Martin Liska <mliska@suse.cz>
4924 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
4925 Do not handle "GNU Pascal".
4926 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
4928 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
4929 from documentation. Likewise.
4930 * dbxout.c (dbxout_range_type): Likewise.
4931 * doc/cpp.texi: Likewise.
4932 * doc/extend.texi: Likewise.
4933 * doc/frontends.texi: Likewise.
4934 * doc/invoke.texi: Remove Pascal entry.
4935 * tree.def (CLEANUP_POINT_EXPR): Likewise.
4936 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
4938 2018-09-25 Martin Liska <mliska@suse.cz>
4941 * doc/invoke.texi: Document all parameters and remove default
4944 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
4947 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
4948 contains HOST_WIDE_INTs when computing its size.
4950 2018-09-24 Jim Wilson <jimw@sifive.com>
4953 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
4955 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
4957 2018-09-24 Andrew Pinski <apinski@marvell.com>
4959 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
4960 access prev before checking it for NULLness in the
4961 AARCH64_FUSE_CMP_BRANCH case.
4963 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
4966 * config/i386/i386.c (rest_of_insert_endbranch): Set
4967 endbr_queued_at_entrance to true and don't insert ENDBR if
4968 x86_function_profiler will be called.
4969 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
4971 * config/i386/i386.h (machine_function): Add
4972 endbr_queued_at_entrance.
4974 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
4976 * genattrtab.c (mk_attr_alt): Use alternative_mask.
4977 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
4979 (check_attr_test): Use alternative_mask.
4980 (get_attr_value): Likewise.
4981 (compute_alternative_mask): Use alternative_mask and XWINT.
4982 (make_alternative_compare): Use alternative_mask.
4983 (attr_alt_subset_p): Use XWINT.
4984 (attr_alt_subset_of_compl_p): Likewise.
4985 (attr_alt_intersection): Use alternative_mask and XWINT.
4986 (attr_alt_union): Likewise.
4987 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
4988 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
4989 (simplify_test_exp): Use alternative_mask and XWINT.
4990 (write_test_expr): Use alternative_mask and XWINT, adjust bit
4991 number calculation to support 64 bits. Generate code that
4992 checks 64-bit masks.
4993 (main): Use alternative_mask.
4994 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
4996 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
4999 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
5000 RETURN+USE when returning via %r14.
5002 2018-09-24 Martin Liska <mliska@suse.cz>
5004 * gcov.c (output_lines): Print colorization legend
5005 for both flag_use_colors and flag_use_hotness_colors.
5008 2018-09-24 Martin Liska <mliska@suse.cz>
5010 * coverage.c (get_coverage_counts): Use warning_at
5011 with current_function_decl location. Use %qD in warning
5014 2018-09-24 Martin Liska <mliska@suse.cz>
5016 * memory-block.h (memory_block_pool::release): Annotate with
5017 valgrind that the memory is not accessible.
5019 2018-09-24 Martin Liska <mliska@suse.cz>
5022 * asan.c: Make asan_handled_variables extern.
5024 * cfgexpand.c (expand_stack_vars): Make sure
5025 a representative is unpoison if another
5026 variable in the partition is handled by
5027 use-after-scope sanitization.
5029 2018-09-24 Richard Biener <rguenther@suse.de>
5031 PR tree-optimization/63155
5032 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
5033 the worklist when the edge of the respective argument isn't
5036 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
5038 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
5039 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
5040 (MASK_CLASS_P): Update for rename.
5041 (MAYBE_MASK_CLASS_P): Ditto.
5042 (REG_CLASS_NAMES): Update.
5043 (REG_CLASS_CONTENT): Update.
5044 * config/i386/i386.c (regclass_map): Update for MASK_REG
5045 and ALL_MASK_REGS rename.
5046 * config/i386/constraints.md (Yk): Update for rename.
5049 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
5051 * config/i386/i386.h (enum reg_class): Remove
5052 EVEX_SSE_REGS and MOD4_SSE_REGS.
5053 (REG_CLASS_NAMES): Update.
5054 (REG_CLASS_CONTENT): Update.
5055 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
5056 registers as ALL_SSE_REGS.
5057 (ix86_additional_allocno_class_p): Remove.
5058 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
5059 (ix86_register_priority): Lower priority of EVEX SSE registers.
5060 Use IN_RANGE macro where appropriate.
5061 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
5062 AVX-5124VNNIW checks.
5063 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
5064 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
5065 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
5066 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
5067 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
5068 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
5069 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
5070 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
5071 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
5072 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
5073 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
5074 * config/i386/constraints.md (Yh): Remove.
5076 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
5078 * config/i386/i386.c (regclass_map): Declare integer REX registers
5081 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
5083 * doc/service.texi (Service): Switch the fsf.org link to https.
5085 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
5088 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5089 Define to speculation_safe_value_not_needed.
5091 2018-09-21 Florian Weimer <fweimer@redhat.com>
5094 * doc/extend.texi (Common Function Attributes): Mention that
5095 noreturn suppresses tail call optimization.
5097 2018-09-21 Jeff Law <law@redhat.com>
5099 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
5100 vr_values::cleanup_edges_and_switches.
5101 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
5103 (identify_jump_threads): Remove EDGE_IGNORE handling.
5104 (execute_vrp): Move handling of to_remove_edges and
5105 to_update_switch_stmts into vr_values class member functions.
5106 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
5107 (to_update_switch_stmts): Likewise.
5108 * vr-values.c: Include cfghooks.h.
5109 (vr_values::vr_values): Initialize to_remove_edges and
5110 to_update_switch_stmts.
5111 (vr_values::~vr_values): Verify to_remove_edges and
5112 to_update_switch_stmts are empty.
5113 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
5114 (vr_values::cleanup_edges_and_switches): New member function.
5115 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
5116 function. Add new data members.
5118 2018-09-21 David Malcolm <dmalcolm@redhat.com>
5120 PR tree-optimization/87309
5121 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
5122 calls with pflags and alt_flags.
5123 (selftest::test_capture_of_dump_calls): Add test of interaction of
5124 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
5126 2018-09-21 Olivier Hainque <hainque@adacore.com>
5128 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
5130 2018-09-21 Olivier Hainque <hainque@adacore.com>
5132 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
5134 2018-09-21 Olivier Hainque <hainque@adacore.com>
5136 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
5137 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
5138 (VXWORKS_LIBS_RTP): Minor reordering.
5140 2018-09-21 Olivier Hainque <hainque@adacore.com>
5142 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
5143 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
5145 2018-09-21 Olivier Hainque <hainque@adacore.com>
5147 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
5148 (PTRDIFF_TYPE): Likewise.
5150 2018-09-21 Olivier Hainque <hainque@adacore.com>
5152 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
5153 triplet, similar to support for VxWorks7.
5154 * config/vxworks-dummy.h: Provide a default definition
5155 of TARGET_VXWORKS64 to 0.
5157 2018-09-21 Olivier Hainque <hainque@adacore.com>
5159 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
5160 * config/vxworks-dummy.h: here.
5162 2018-09-21 Olivier Hainque <hainque@adacore.com>
5164 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
5166 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
5167 Bo Zhou <zbo.zhou@hisilicon.com>
5169 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
5170 * config/aarch64/aarch64-tune.md: Regenerated.
5171 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
5172 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
5173 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
5175 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
5176 Julian Brown <julian@codesourcery.com>
5178 * builtins.c (get_builtin_sync_mem): Handle address spaces.
5180 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
5182 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
5183 if the call takes a static chain.
5185 2018-09-21 Martin Liska <mliska@suse.cz>
5187 * auto-profile.c (autofdo_source_profile::read): Do not
5189 (read_profile): Do not add working sets.
5190 (read_autofdo_file): Remove sum_all.
5191 (afdo_callsite_hot_enough_for_early_inline): Remove const
5193 * coverage.c (struct counts_entry): Remove gcov_summary.
5194 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
5195 do not support GCOV_TAG_PROGRAM_SUMMARY.
5196 (get_coverage_counts): Remove summary and expected
5198 * coverage.h (get_coverage_counts): Likewise.
5199 * doc/gcov-dump.texi: Remove -w option.
5200 * gcov-dump.c (dump_working_sets): Remove.
5201 (main): Do not support '-w' option.
5202 (print_usage): Likewise.
5203 (tag_summary): Likewise.
5204 * gcov-io.c (gcov_write_summary): Do not dump
5206 (gcov_read_summary): Likewise.
5207 (gcov_histo_index): Remove.
5208 (gcov_histogram_merge): Likewise.
5209 (compute_working_sets): Likewise.
5210 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
5212 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
5213 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
5214 (GCOV_HISTOGRAM_SIZE): Remove.
5215 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
5216 (struct gcov_summary): Simplify rapidly just
5217 to runs and sum_max fields.
5218 (gcov_histo_index): Remove.
5219 (NUM_GCOV_WORKING_SETS): Likewise.
5220 (compute_working_sets): Likewise.
5221 * gcov-tool.c (print_overlap_usage_message): Remove
5222 trailing empty line.
5223 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
5224 (output_lines): Remove program related line.
5225 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
5226 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
5228 (input_profile_summary): Do not read it.
5229 (merge_profile_summaries): And do not merge it.
5230 (input_symtab): Do not call removed function.
5231 * modulo-sched.c (sms_schedule): Do not print sum_max.
5232 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
5233 removed when histogram method was invented.
5234 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
5236 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
5238 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
5239 and dump selected value.
5240 * profile.c (add_working_set): Remove.
5241 (get_working_sets): Likewise.
5242 (find_working_set): Likewise.
5243 (get_exec_counts): Do not work with working sets.
5244 (read_profile_edge_counts): Do not inform as sum_max is removed.
5245 (compute_branch_probabilities): Likewise.
5246 (compute_value_histograms): Remove argument for call of
5247 get_coverage_counts.
5248 * profile.h: Do not make gcov_summary const.
5250 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
5252 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
5254 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
5256 PR tree-optimization/86990
5257 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
5258 Check that the entire merged store group is made of constants only for
5261 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
5263 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
5264 (VTABLE_VERIFICATION_SPEC): Likewise.
5265 (SANITIZER_EARLY_SPEC): Likewise.
5266 (SANITIZER_SPEC): Likewise.
5267 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
5268 * doc/invoke.texi (Link Options): Document -r.
5270 2018-09-20 Richard Biener <rguenther@suse.de>
5273 * gimplify.c (gimplify_expr): Retain alignment of
5274 addressable lvalue in dereference.
5276 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
5279 * configure.ac: Check for .loc is_stmt support.
5280 * configure, config.in: Rebuilt.
5281 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
5284 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
5286 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
5287 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
5290 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
5292 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5294 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
5297 2018-09-20 Marek Polacek <polacek@redhat.com>
5299 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
5301 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
5303 PR tree-optimization/87288
5304 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
5305 into account when determining PEELING_FOR_NITERS.
5307 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
5309 PR tree-optimization/86877
5310 * tree-vect-loop.c (vect_analyze_loop_2): Call
5311 vect_verify_datarefs_alignment.
5313 2018-09-19 Marek Polacek <polacek@redhat.com>
5315 * doc/invoke.texi: Document -Wclass-conversion.
5317 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
5319 * config/pa/pa.c (pa_adjust_priority): Delete.
5320 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
5322 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
5323 (atomic_storehi): Likewise.
5324 (atomic_storesi): Likewise.
5325 (atomic_loaddi): Restore compare and swap exchange loop code.
5327 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
5329 PR rtl-optimization/86902
5330 * combine.c (try_combine): When changing the CC mode used, don't change
5331 an unrelated mode in other_insn to that new CC mode.
5333 2018-09-19 David Malcolm <dmalcolm@redhat.com>
5335 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
5336 with %T in place of calls to dump_generic_expr.
5337 (prune_runtime_alias_test_list): Likewise.
5338 (create_runtime_alias_checks): Likewise.
5339 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
5340 (vect_analyze_data_ref_dependence): Likewise.
5341 (vect_slp_analyze_data_ref_dependence): Likewise.
5342 (vect_record_base_alignment): Likewise. Use %G in place of call
5343 to dump_gimple_stmt.
5344 (vect_compute_data_ref_alignment): Likewise.
5345 (verify_data_ref_alignment): Likewise.
5346 (vect_find_same_alignment_drs): Likewise.
5347 (vect_analyze_group_access_1): Likewise.
5348 (vect_analyze_data_ref_accesses): Likewise.
5349 (dependence_distance_ge_vf): Likewise.
5350 (dump_lower_bound): Likewise.
5351 (vect_prune_runtime_alias_test_list): Likewise.
5352 (vect_find_stmt_data_reference): Likewise.
5353 (vect_analyze_data_refs): Likewise.
5354 (vect_create_addr_base_for_vector_ref): Likewise.
5355 (vect_create_data_ref_ptr): Likewise.
5356 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
5357 (vect_can_advance_ivs_p): Likewise.
5358 (vect_update_ivs_after_vectorizer): Likewise.
5359 (vect_gen_prolog_loop_niters): Likewise.
5360 (vect_prepare_for_masked_peels): Likewise.
5361 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
5362 (vect_determine_vectorization_factor): Likewise.
5363 (vect_is_simple_iv_evolution): Likewise.
5364 (vect_analyze_scalar_cycles_1): Likewise.
5365 (vect_analyze_loop_operations): Likewise.
5366 (report_vect_op): Likewise.
5367 (vect_is_slp_reduction): Likewise.
5368 (check_reduction_path): Likewise.
5369 (vect_is_simple_reduction): Likewise.
5370 (vect_create_epilog_for_reduction): Likewise.
5371 (vect_finalize_reduction:): Likewise.
5372 (vectorizable_induction): Likewise.
5373 (vect_transform_loop_stmt): Likewise.
5374 (vect_transform_loop): Likewise.
5375 (optimize_mask_stores): Likewise.
5376 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
5377 (vect_split_statement): Likewise.
5378 (vect_recog_over_widening_pattern): Likewise.
5379 (vect_recog_average_pattern): Likewise.
5380 (vect_determine_min_output_precision_1): Likewise.
5381 (vect_determine_precisions_from_range): Likewise.
5382 (vect_determine_precisions_from_users): Likewise.
5383 (vect_mark_pattern_stmts): Likewise.
5384 (vect_pattern_recog_1): Likewise.
5385 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5386 (vect_record_max_nunits): Likewise.
5387 (vect_build_slp_tree_1): Likewise.
5388 (vect_build_slp_tree_2): Likewise.
5389 (vect_print_slp_tree): Likewise.
5390 (vect_analyze_slp_instance): Likewise.
5391 (vect_detect_hybrid_slp_stmts): Likewise.
5392 (vect_detect_hybrid_slp_1): Likewise.
5393 (vect_slp_analyze_operations): Likewise.
5394 (vect_slp_analyze_bb_1): Likewise.
5395 (vect_transform_slp_perm_load): Likewise.
5396 (vect_schedule_slp_instance): Likewise.
5397 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
5398 (vect_mark_stmts_to_be_vectorized): Likewise.
5399 (vect_init_vector_1): Likewise.
5400 (vect_get_vec_def_for_operand): Likewise.
5401 (vect_finish_stmt_generation_1): Likewise.
5402 (vect_check_load_store_mask): Likewise.
5403 (vectorizable_call): Likewise.
5404 (vectorizable_conversion): Likewise.
5405 (vectorizable_operation): Likewise.
5406 (vectorizable_load): Likewise.
5407 (vect_analyze_stmt): Likewise.
5408 (vect_is_simple_use): Likewise.
5409 (vect_get_vector_types_for_stmt): Likewise.
5410 (vect_get_mask_type_for_stmt): Likewise.
5411 * tree-vectorizer.c (increase_alignment): Likewise.
5413 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
5415 * doc/rtl.texi: Adjust vec_select description.
5416 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
5417 non-constant selectors.
5419 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
5421 * config/aarch64/aarch64-protos.h
5422 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
5423 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
5424 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
5425 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
5426 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
5427 (AARCH64_FL_PROFILE): Move index so flags are ordered.
5428 (AARCH64_ISA_RCPC8_4): New flag.
5429 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
5430 to aarch64_offset_9bit_signed_unscaled_p.
5431 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
5433 * config/aarch64/constraints.md (Ust): New constraint.
5434 * config/aarch64/predicates.md.
5435 (aarch64_9bit_offset_memory_operand): New predicate.
5436 (aarch64_rcpc_memory_operand): New predicate.
5438 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
5440 PR rtl-optimization/87361
5441 * rtlanal.c (nonzero_bits1): Revert accidental change.
5443 2018-09-19 Richard Biener <rguenther@suse.de>
5445 PR tree-optimization/87349
5446 PR tree-optimization/87342
5447 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
5449 2018-09-18 Marek Polacek <polacek@redhat.com>
5451 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
5452 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
5454 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
5456 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
5458 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
5460 PR rtl-optimization/86882
5461 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
5463 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
5465 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
5466 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
5468 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
5471 * doc/invoke.texi (Link Options): Fix formatting and grammar.
5473 2018-09-18 Richard Biener <rguenther@suse.de>
5476 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
5477 (compute_samebase_partition_bases): Likewise.
5478 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
5479 (gimple_can_coalesce_p): Simplify.
5481 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
5483 Handle a library implementation of ffs calling __builtin_ffs.
5484 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
5485 (mmix_init_libfuncs): New function: make __builtin_ffs expand
5488 2018-09-17 David Malcolm <dmalcolm@redhat.com>
5490 * diagnostic-show-locus.c (class layout_range): Add field
5492 (layout_range::layout_range): Add "original_idx" param and use it
5493 to initialize new field.
5494 (make_range): Use 0 for original_idx.
5495 (layout::layout): Pass in index to calls to
5496 maybe_add_location_range.
5497 (layout::maybe_add_location_range): Add param "original_idx" and
5498 pass it on to layout_range.
5499 (layout::print_any_labels): Pass on range->m_original_idx to
5501 (gcc_rich_location::add_location_if_nearby): Use 0 for
5503 * gcc-rich-location.h (text_range_label::get_text): Update for new
5505 (range_label_for_type_mismatch::get_text): Likewise.
5507 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
5509 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
5511 2018-09-17 David Malcolm <dmalcolm@redhat.com>
5513 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
5514 format_string_diagnostic_t.
5515 (fmtwarn_n): Likewise.
5516 * substring-locations.c
5517 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
5518 (format_warning_n_va): Convert to...
5519 (format_string_diagnostic_t::emit_warning_n_va): ...this.
5520 (format_warning_va): Convert to...
5521 (format_string_diagnostic_t::emit_warning_va): ...this.
5522 (format_warning_at_substring): Convert to...
5523 (format_string_diagnostic_t::emit_warning): ...this.
5524 (format_warning_at_substring_n): Convert to...
5525 (format_string_diagnostic_t::emit_warning_n): ...this.
5526 * substring-locations.h (class format_string_diagnostic_t): New
5528 (format_warning_va): Convert to
5529 format_string_diagnostic_t::emit_warning_va.
5530 (format_warning_n_va): Convert to
5531 format_string_diagnostic_t::emit_warning_n_va.
5532 (format_warning_at_substring): Convert to
5533 format_string_diagnostic_t::emit_warning.
5534 (format_warning_at_substring_n): Convert to
5535 format_string_diagnostic_t::emit_warning_n.
5537 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
5538 Bernd Schmidt <bernds_cb1@t-online.de>
5540 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
5543 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
5545 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
5546 operand 0 predicate to nonimmediate operand.
5547 (rint<mode>2_frndint): Remove insn pattern.
5548 (rint<mode>2): Change operand 1 predicate to general_operand.
5549 Extend operand 1 to XFmode and generate rintxf2 insn.
5550 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
5551 Do not use X87MODEF mode macro.
5552 (frndintxf2_<rounding>_i387): Rename from
5553 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
5554 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
5555 to XFmode and generate significandxf3 insn.
5557 2018-09-17 Richard Biener <rguenther@suse.de>
5559 PR tree-optimization/87328
5560 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
5561 visiting unexecutable backedges when not iterating.
5562 (do_rpo_vn): Mark all edges not executable even when not
5565 2018-09-17 Martin Jambor <mjambor@suse.cz>
5568 * doc/invoke.texi (Warning Options): Likewise.
5570 2018-09-17 Richard Biener <rguenther@suse.de>
5572 PR tree-optimization/87301
5573 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
5574 clean EH info from leftover copy assignments.
5576 2018-09-17 Martin Liska <mliska@suse.cz>
5578 PR gcov-profile/85871
5579 * gcov.c (output_intermediate_file): Fix out of bounds
5582 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
5584 * config/arc/arc.c: Object attributes for core4 not reflected
5586 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
5589 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
5591 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
5593 2018-09-17 Martin Liska <mliska@suse.cz>
5595 * doc/gcov.texi: Document new option --use-hotness-colors.
5596 * gcov.c (struct source_info): Declare new field.
5597 (source_info::source_info): Set default for maximum_count.
5598 (print_usage): Add new -q option.
5599 (process_args): Process it.
5600 (accumulate_line_info): Save src->maximum_count.
5601 (output_line_beginning): Make color line number if
5602 flag_use_hotness_colors is set.
5603 (output_line_details): Pass default argument value.
5604 (output_lines): Pass src->maximum_count.
5606 2018-09-17 Martin Liska <mliska@suse.cz>
5608 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
5609 Use processor_names table.
5610 * config/i386/i386.c (ix86_default_align): Use
5611 processor_cost_table for alignment values.
5612 (ix86_option_override_internal): Use processor_names.
5613 (ix86_function_specific_print): Likewise.
5614 * config/i386/i386.h (struct processor_costs):
5615 Add alignment values.
5616 (struct ptt): Remove and replace with const char *.
5617 * config/i386/x86-tune-costs.h (struct processor_costs):
5618 Declare default alignments for all costs.
5620 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
5622 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
5623 symbolics or VR_VARYING ranges for ABS_EXPR.
5624 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
5625 when range will wrap.
5627 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
5630 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
5631 before and after a JUMP_TABLE_DATA.
5633 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
5636 * dojump.c (do_compare_and_jump): Canonicalize function pointers
5637 when one operand is a function pointer. Use POINTER_TYPE_P and
5638 FUNC_OR_METHOD_TYPE_P.
5639 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
5640 * fold-const.c (build_range_check): Likewise.
5641 * match.pd (simple_comparison): Likewise.
5643 2018-09-14 David Malcolm <dmalcolm@redhat.com>
5646 * spellcheck.c (get_edit_distance_cutoff): New function.
5647 (selftest::test_edit_distance_unit_test_oneway): Rename to...
5648 (selftest::test_get_edit_distance_one_way): ...this.
5649 (selftest::test_get_edit_distance_unit): Rename to...
5650 (selftest::test_get_edit_distance_both_ways): ...this.
5651 (selftest::test_edit_distances): Move tests to this new function,
5652 and test some more pairs of strings. Update for above renaming.
5653 (selftest::get_old_cutoff): New function.
5654 (selftest::test_get_edit_distance_cutoff): New function.
5655 (selftest::assert_suggested_for): New function.
5656 (ASSERT_SUGGESTED_FOR): New macro.
5657 (selftest::assert_not_suggested_for): New function.
5658 (ASSERT_NOT_SUGGESTED_FOR): New macro.
5659 (selftest::test_suggestions): New function.
5660 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
5661 tests to selftest::test_edit_distances and call it. Add calls to
5662 selftest::test_get_edit_distance_cutoff and
5663 selftest::test_suggestions.
5664 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
5665 (best_match::consider): Replace hard-coded cutoff calculation with
5667 (best_match::get_cutoff): New declaration.
5668 (best_match::get_best_meaningful_candidate): Likewise.
5670 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
5672 * builtins.c (fold_builtin_strlen): Remove TODO comment.
5674 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
5677 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
5679 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
5680 terminated string literal.
5682 2018-09-14 Martin Sebor <msebor@redhat.com>
5684 * builtins.c (unterminated_array): Handle ARRAY_REF.
5685 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
5686 * builtins.h (unterminated_array): Declare extern.
5687 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
5689 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
5692 2018-09-14 Martin Sebor <msebor@redhat.com>
5693 Jeff Law <law@redhat.com>
5695 * builtins.c (unterminated_array): New.
5696 (expand_builtin_strcpy): Adjust.
5697 (expand_builtin_strcpy_args): Detect unterminated arrays.
5698 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
5699 unterminated arrays.
5700 * gimple-fold.h (get_maxval_strlen): Add argument.
5701 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
5703 * gimple-fold.c (get_range_strlen): Add argument.
5704 (get_maxval_strlen): Adjust.
5705 * gimple-fold.h (get_range_strlen): Add argument.
5707 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
5709 * config/i386/movdirintrin.h: Fix copyright year.
5711 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
5713 * reg-stack.c: Include regs.h.
5714 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
5715 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
5716 FIRST_STACK_REG, not DFmode.
5717 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
5718 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
5719 (change stack): Default register mode to the reg_raw_mode of
5720 FIRST_STACK_REG, not DFmode.
5721 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
5722 (*swapxf): Rename from swapxf.
5724 2018-09-14 Carl Love <cel@us.ibm.com>
5726 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
5727 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
5729 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
5732 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
5735 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
5738 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
5740 2018-09-14 Jason Merrill <jason@redhat.com>
5742 Fix --enable-gather-detailed-mem-stats.
5743 * hash-table.c (hash_table_usage): Change from variable to function.
5744 * hash-table.h: Adjust.
5745 * Makefile.in: Add missing dependencies on hash-table.h.
5747 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5749 PR tree-optimization/87259
5751 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
5752 execute_cse_reciprocals_1 has tried transforming.
5754 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
5756 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
5757 VR_VARYING for PLUS/MINUS_EXPR.
5759 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
5761 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
5764 2018-09-14 Richard Biener <rguenther@suse.de>
5767 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
5768 bits for the merged partition.
5770 2018-09-13 Martin Sebor <msebor@redhat.com>
5771 Bernd Edlinger <bernd.edlinger@hotmail.de>
5773 * builtins.h (c_srlen): Add argument.
5774 * builtins.c (warn_string_no_nul): New function.
5775 (c_strlen): Add argument and use it. Update recursive calls.
5776 Pass DECL argument to string_constant to get info on non
5777 terminated strings. Update *NONSTR as needed.
5778 (fold_builtin_strlen): Add argument to calls to c_strlen.
5779 Warn for unterminated arrays.
5780 (warn_string_no_null): Add prototype.
5781 * expr.c (string_constant): Update arguments. Update recursive
5782 calls appropriately. Detect missing NUL terminator and outermost
5783 declaration its missing in.
5784 Improve checks for arrays with nonzero lower bound or elements
5785 that are not a single byte. Simplify offset computation.
5786 Simplify checks for non-NUL terminated strings.
5787 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
5788 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
5790 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
5792 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
5794 * fold-const.c (c_getstr): Fix function comment. Remove unused third
5795 argument. Fix range checks.
5796 * fold-const.h (c_getstr): Adjust protoype.
5797 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
5798 string is constant but contains no NUL byte.
5800 * expr.c (string_constant): Adjust function comment.
5801 Remove bogus check for zero termination.
5803 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
5805 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
5806 (get_constant_size): Don't make STRING_CSTs larger than they are.
5807 (check_string_literal): New check function for STRING_CSTs.
5808 (output_constant): Use it.
5810 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
5813 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
5815 2018-09-13 Richard Biener <rguenther@suse.de>
5817 PR tree-optimization/87263
5818 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
5819 (struct unwind_state): Add max_rpo field.
5820 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
5821 Compute max_rpo, the max RPO number a block can be backwards reached
5822 from. Re-write non-iterating mode to a RPO ordered worklist approach,
5823 separating it from the iterating mode.
5825 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
5827 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
5828 (rfs_decision): New scheduling decision.
5830 2018-09-13 Richard Biener <rguenther@suse.de>
5833 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
5834 (vn_nary_op_insert_pieces_predicated): Do not write useless
5835 valid_dominated_by_p entry outside of the allocated storage.
5837 2018-09-13 Omar Sandoval <osandov@osandov.com>
5838 Tom de Vries <tdevries@suse.de>
5841 * dwarf2out.c (is_c): New function.
5842 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
5844 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
5847 * config/aarch64/aarch64.md (*aarch64_bfxil):
5849 * config/aarch64/constraints.md (Ulc): Define.
5850 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
5852 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
5855 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
5857 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
5858 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
5859 aarch64_layout_frame call.
5860 (aarch64_expand_epilogue): Likewise.
5861 (aarch64_initial_elimination_offset): Likewise.
5862 (aarch64_get_separate_components): Likewise.
5863 (aarch64_use_return_insn_p): Likewise.
5864 (aarch64_layout_frame): Remove unneeded check.
5866 2018-09-13 Jakub Jelinek <jakub@redhat.com>
5868 * configure.ac: Only append
5869 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
5870 gcc_config_arguments if it was never reconfigured or last reconfigure
5871 was with different arguments.
5872 * configure: Regenerated.
5874 2018-09-13 Jakub Jelinek <jakub@redhat.com>
5875 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5878 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
5879 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
5881 2018-09-13 Jakub Jelinek <jakub@redhat.com>
5883 PR tree-optimization/87287
5884 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
5885 X % C == 0 to X % (unsigned) C == 0 optimization to ...
5886 * match.pd (X % C == 0): ... here. New optimization.
5888 2018-09-12 Jakub Jelinek <jakub@redhat.com>
5891 * expr.h (maybe_optimize_mod_cmp): Declare.
5892 * expr.c (mod_inv): New function.
5893 (maybe_optimize_mod_cmp): New function.
5894 (do_store_flag): Use it.
5895 * cfgexpand.c (expand_gimple_cond): Likewise.
5897 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
5898 Julian Brown <julian@codesourcery.com>
5901 * gimplify.c (gimplify_scan_omp_clauses): Set
5902 target_firstprivatize_array_bases in OpenACC parallel and kernels
5903 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
5904 OpenACC data regions.
5906 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
5908 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
5909 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
5911 2018-09-12 Richard Biener <rguenther@suse.de>
5913 PR tree-optimization/87280
5914 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
5915 edge but unreachable target.
5916 (do_rpo_vn): For conservatively handling a PHI only mark
5917 the backedge executable but not the block reachable.
5919 2018-09-12 Richard Biener <rguenther@suse.de>
5921 PR tree-optimization/87266
5922 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
5925 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
5927 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
5929 ("trunc<BFP:mode><DFP_ALL:mode>2")
5930 ("trunc<DFP_ALL:mode><BFP:mode>2")
5931 ("extend<BFP:mode><DFP_ALL:mode>2")
5932 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
5933 according to the target operand type.
5935 2018-09-12 Jakub Jelinek <jakub@redhat.com>
5936 Andreas Krebbel <krebbel@linux.ibm.com>
5938 PR tree-optimization/86844
5939 * gimple-ssa-store-merging.c
5940 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
5941 there are any overlapping stores in between them, make sure they are
5942 also coalesced or we give up completely.
5944 2018-09-12 Jakub Jelinek <jakub@redhat.com>
5947 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
5948 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
5950 2018-09-12 Tom de Vries <tdevries@suse.de>
5952 * common.opt (gdescribe-dies): Add option.
5953 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
5954 attribute for artifical and nameless decls.
5955 (dwarf2out_register_external_die): Add description attribute to
5956 external reference die.
5957 (add_desc_attribute): New functions.
5958 (gen_subprogram_die): Add description attribute to
5959 DW_TAG_call_site_parameter.
5960 * tree-pretty-print.c (print_generic_expr_to_str): New function.
5961 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
5962 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
5964 (@item -gdescribe-dies): Add.
5966 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
5968 * tree-vrp.c (vrp_shift_undefined_p): Remove.
5969 (extract_range_from_binary_expr_1: Call
5970 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
5971 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
5974 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
5976 * gimple-ssa-warn-alloca.c
5977 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
5978 field for ALLOCA_BOUND_*_LARGE.
5980 2018-09-11 Nathan Sidwell <nathan@acm.org>
5982 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
5984 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
5986 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
5987 for clobbers. Remove obsolete comment.
5989 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
5991 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
5992 mpxchk, mpxld and mpxst types.
5993 (define_attr length_immediate): Remove all processing of mpx types.
5994 (define_attr prefix_0f): Ditto.
5995 (define_attr memory): Ditto.
5997 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
5999 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
6000 (log<mode>2): Change operand 1 predicate to general_operand.
6001 Extend operand 1 to XFmode and generate logxf3 insn.
6002 (log10<mode>2): Change operand 1 predicate to general_operand.
6003 Extend operand 1 to XFmode and generate log10xf3 insn.
6004 (log2<mode>2): Change operand 1 predicate to general_operand.
6005 Extend operand 1 to XFmode and generate log2xf3 insn.
6006 (fyl2xp1_extend<mode>xf3_i387): Remove.
6007 (log1p<mode>2): Change operand 1 predicate to general_operand.
6008 Extend operand 1 to XFmode and generate log1pxf3 insn.
6009 (fxtract_extend<mode>xf3_i387): Remove.
6010 (logb<mode>2): Change operand 1 predicate to general_operand.
6011 Extend operand 1 to XFmode and generate logbxf3 insn.
6012 (ilogb<mode>2): Change operand 1 predicate to general_operand.
6013 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
6014 (significand<mode>2): Change operand 1 predicate to general_operand.
6015 Extend operand 1 to XFmode and generate significandxf3 insn.
6017 2018-09-11 Nathan Sidwell <nathan@acm.org>
6019 * gcc.c (perror_with_name, pfatal_with_name): Delete.
6020 (load_specs): Use fatal_error.
6021 (DELETE_IF_ORDINARY, process_command): Use error.
6022 (execute, run_attempt): Use fatal_error.
6024 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
6026 * diagnostic-core.h (sorry_at): New prototype.
6027 * diagnostic.c (sorry_at): New function.
6029 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
6031 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
6032 by zero as VR_UNDEFINED.
6034 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
6036 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
6037 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
6038 (<sincos>mode2): New expander.
6039 (sincos_extend<mode>xf3_i387): Remove insn pattern.
6040 (sincos -> sin, cos splitters): Remove splitter patterns.
6041 (sincos<mode>3): Change operand 2 predicate to general_operand.
6042 Extend operand 2 to XFmode and generate sincosxf3 insn.
6043 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
6044 Change operand 3 predicate to const1_operand.
6045 (fptan_extend<mode>xf4_i387): Remove insn pattern.
6046 (tanxf2): Update operands in the call to fptanxf4_i387.
6047 (tan<mode>2): Change operand 1 predicate to general_operand.
6048 Extend operand 1 to XFmode and generate tanxf3 insn.
6049 (atan2xf3): Rename from *fpatanxf3_i387.
6050 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
6051 (atan2xf3): Remove expander.
6052 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
6053 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
6054 (atan<mode>2): Change operand 1 predicate to general_operand.
6055 Extend operand 1 to XFmode and generate atanxf3 insn.
6057 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
6059 * config/i386/i386.md (x87/SSE constant load splitter): Use
6060 memory_operand instead of nonimmediate_operand for input operand
6063 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
6065 * config/i386/i386.md (float partial SSE register stall splitter): Move
6066 splitter near its instruction pattern.
6067 (float_extend partial SSE register stall splitter): Ditto.
6068 (float_truncate partial SSE register stall splitter): Ditto.
6070 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
6073 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6074 to speculation_safe_value_not_needed.
6077 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
6078 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
6079 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
6080 leaf_function_p, instead use has_hard_reg_initial_val.
6082 2018-09-09 Nathan Sidwell <nathan@acm.org>
6084 * gcc.h (pfatal_with_name): Don't declare here.
6085 * gcc.c (pfatal_with_name): Make static.
6087 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
6089 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
6092 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
6094 PR rtl-optimization/85458
6095 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
6096 priority hook to reduce the priority of EXPR.
6098 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
6100 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
6101 DImode for x87 on 32bit targets. Conditionally disable x87 modes
6102 with X87_ENABLE_FLOAT. Remove preparation code.
6103 (*float<SWI48:mode><MODEF:mode>2): Rename from
6104 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
6105 math using "enabled" attribute.
6106 (*floatdi<MODEF:mode>2_i387): Rename from
6107 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
6108 enable for 32bit targets only.
6109 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
6111 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
6112 as operand 1 predicate. Rewrite as define_insn_and_split.
6113 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
6115 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
6117 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
6118 to fallthru to FLOAT case.
6120 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
6123 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
6124 around folding of vec_sl to handle out of range shift values.
6126 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
6128 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
6129 Update callers to gen_fix_trunc<mode>_i387_fisttp
6130 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
6131 nonimmediate_operand.
6132 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
6133 and corresponding splitters.
6134 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
6135 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
6136 (fix_truncdi_i387_with_temp): Remove insn pattern
6137 and corresponding splitters.
6138 (fix_trunc<mode>_i387): Change operand 0 predicate to
6139 nonimmediate_operand.
6140 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
6141 and corresponding splitters.
6142 (*fistdi2_1): Remove.
6144 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
6145 (lrintxfdi2): Remove expander. Reimplement as define_insn.
6146 (*fist<mode>2_1): Remove.
6147 (fist<mode>2): Ditto.
6148 (fist<mode>2_with_temp): Remove insn pattern and corresponding
6150 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
6151 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
6152 (fistdi2_<rounding>): Change operand 0 predicate to
6153 nonimmediate_operand.
6154 (fistdi2_<rounding>_with_temp): Remove insn pattern
6155 and corresponding splitters.
6156 (fist<mode>2_<rounding>): Change operand 0 predicate to
6157 nonimmediate_operand.
6158 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
6159 and corresponding splitters.
6161 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
6163 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
6165 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
6168 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
6170 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
6171 early gimple folding of vec_splat().
6172 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
6173 * gimple-fold.h: Add an extern define for tree_vec_extract().
6175 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
6177 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
6178 wrappers around TREE_TYPE comparisons.
6180 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
6183 * config/s390/predicates.md: Add nonsym_memory_operand.
6184 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
6185 contains a SYMBOL_REF, load it into an intermediate pseudo.
6186 (s390_emit_compare_and_swap): Legitimize operand.
6187 * config/s390/s390.md: Use the new nonsym_memory_operand
6188 with UNSPECV_CAS patterns.
6190 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
6193 * config/s390/s390-passes.def: New file.
6194 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
6196 (make_pass_s390_early_mach): Add declaration.
6197 * config/s390/s390.c (make_pass_s390_early_mach):
6198 (s390_option_override): Remove dynamic registration.
6199 * config/s390/t-s390: Add s390-passes.def.
6201 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
6203 * config/s390/s390.c (s390_decompose_constant_pool_ref):
6204 Remove UNSPEC_LTREL_BASE check.
6205 (annotate_constant_pool_refs): Likewise.
6206 (find_constant_pool_ref): Likewise.
6207 (find_ltrel_base): Removed.
6208 (replace_ltrel_base): Removed.
6209 (s390_mainpool_finish): Remove replace_ltrel_base call.
6210 (s390_chunkify_start): Remove pending LTREL_BASE logic.
6211 (s390_chunkify_finish): Remove replace_ltrel_base call.
6212 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
6214 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
6217 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6218 to speculation_safe_value_not_needed.
6220 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
6221 Bernd Schmidt <bernds_cb1@t-online.de>
6223 * config/nvptx/nvptx-opts.h: New file.
6224 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
6225 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
6227 (TARGET_SM35): New macro.
6228 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
6230 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
6232 (misa=): New option.
6233 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
6235 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
6237 * config/i386/i386.md (truncdfsf2): Remove expander.
6238 (truncdfsf2_with_temp): Ditto.
6239 (truncxf<mode>2): Ditto.
6240 (*truncdfsf_fast_mixed): Remove insn pattern.
6241 (*truncdfsf_fast_i387): Ditto.
6242 (*truncdfsf_mixed): Ditto.
6243 (*truncdfsf_i387): Ditto.
6244 (*truncdfsf2_i387_1): Ditto.
6245 (*truncxfsf2_mixed): Ditto.
6246 (*truncxfdf2_mixed): Ditto.
6247 (*truncxf<mode>2_i387_noop): Ditto. Update callers
6248 to call gen_truncxf<mode>2 instead.
6249 (*truncxf<mode>2_i387): Remove.
6250 (reg->reg splitters): Remove splitter pattern.
6251 (reg->mem splitters): Ditto.
6253 (truncdfsf2): New insn pattern.
6254 (truncxf<mode>2): Ditto.
6256 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6258 * tree-ssa-math-opts.c (is_mult_by): New function.
6259 (is_square_of): Use the above.
6260 (optimize_recip_sqrt): New function.
6261 (pass_cse_reciprocals::execute): Use the above.
6263 2018-09-05 Richard Biener <rguenther@suse.de>
6266 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
6267 to zero-init the emplaced vec.
6269 2018-09-05 Martin Liska <mliska@suse.cz>
6271 PR tree-optimization/87205
6272 * tree-switch-conversion.c (pass_lower_switch::execute):
6273 Group cases for switch statements.
6275 2018-09-05 Richard Biener <rguenther@suse.de>
6277 PR tree-optimization/87217
6278 * tree-ssa-sccvn.c (vuse_valueize): New.
6279 (vn_reference_lookup_pieces): Use it.
6280 (vn_reference_lookup): Likewise.
6282 2018-09-05 Nathan Sidwell <nathan@acm.org>
6285 * stor-layout.c (place_field): Scan forwards to check last
6286 bitfield when ms_bitfield_placement is in effect.
6288 2018-09-05 Richard Biener <rguenther@suse.de>
6291 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
6294 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
6295 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
6297 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
6298 * config.gcc (extra_objs): Build it.
6299 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
6301 * config/aarch64/aarch64-passes.def
6302 (pass_tag_collision_avoidance): New pass.
6303 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
6304 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
6305 (aarch64_classify_address): Remove static qualifier.
6306 (aarch64_address_info, aarch64_address_type): Move to...
6307 * config/aarch64/aarch64-protos.h: ... here.
6308 (make_pass_tag_collision_avoidance): New function.
6309 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
6312 2018-09-05 Martin Liska <mliska@suse.cz>
6314 * doc/gcov.texi: Update documentation of humar
6316 * gcov.c (format_count): Print one decimal place, it provides
6317 more fine number of situations like '1G' vs. '1.4G'.
6319 2018-09-05 Martin Liska <mliska@suse.cz>
6322 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
6323 * optc-gen.awk: Allow 'Var' for Deprecated options in order
6324 to generate a MASK value.
6326 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
6329 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
6330 if frame pointer isn't used.
6331 (compute_frame_pointer_to_fb_displacement): Likewise.
6333 2018-09-04 Jakub Jelinek <jakub@redhat.com>
6336 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
6337 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
6338 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
6339 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
6340 and OPTION_MASK_ISA_XSAVEC_UNSET.
6342 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
6344 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
6345 XOR operations in NAND case.
6347 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
6349 * wide-int-range.cc (wide_int_range_convert): New.
6350 * wide-int-range.h (wide_int_range_convert): New.
6351 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
6352 code into wide_int_range_convert.
6353 (extract_range_into_wide_ints): Do not munge anti range constants
6354 into the entire domain. Just return the range back.
6356 2018-09-04 Martin Liska <mliska@suse.cz>
6358 * genmatch.c (output_line_directive): Add new argument
6360 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
6362 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
6364 * doc/invoke.texi (Option Summary): Add whitespace.
6366 * doc/invoke.texi (Option Summary): Add -Waligned-new.
6368 2018-09-04 Richard Biener <rguenther@suse.de>
6370 PR tree-optimization/87211
6371 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
6372 backedge value we're supposed to treat as VARYING also number
6373 the PHI to VARYING in case it got a different value-number already.
6375 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
6377 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
6378 (extract_range_from_binary_expr_1): Do not call
6379 vrp_can_optimize_bit_op.
6380 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
6382 (wide_int_range_get_mask_and_bounds): New.
6383 (wide_int_range_optimize_bit_op): New.
6384 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
6385 (wide_int_range_bit_and): Same.
6386 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
6387 (wide_int_range_optimize_bit_op): New.
6388 (wide_int_range_get_mask_and_bounds): New.
6390 2018-09-04 Richard Biener <rguenther@suse.de>
6392 PR tree-optimization/87176
6393 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
6394 variable. When value-numbering a virtual PHI node make sure
6395 to not value-number to the backedge value.
6397 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
6399 * doc/extend.texi (Long Long, Hex Floats): Document support for
6400 long long and hex floats in more recent versions of ISO C++.
6402 2018-09-03 Richard Biener <rguenther@suse.de>
6404 PR tree-optimization/87177
6405 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
6408 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
6410 * bb-reorder.c (edge_order): Convert to C-qsort-style
6411 tri-state comparator.
6412 (reorder_basic_blocks_simple): Change std::stable_sort to
6415 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
6417 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
6418 tri-state comparator.
6419 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
6421 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
6423 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
6424 (mergesort): ... here as maximum count for using netsort.
6425 (gcc_qsort): Set nlim to 3 if stable sort is requested.
6426 (gcc_stablesort): New.
6427 * system.h (gcc_stablesort): Declare.
6429 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
6431 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
6432 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
6433 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
6435 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
6437 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
6438 lxsdx and stxsdx alternatives.
6439 (*mov<mode>_hardfloat64): Ditto.
6440 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
6442 2018-09-03 Richard Biener <rguenther@suse.de>
6444 PR tree-optimization/87200
6445 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
6448 2018-09-03 Martin Liska <mliska@suse.cz>
6450 PR tree-optimization/87201
6451 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
6452 Fix parenthesis in an expression.
6454 2018-09-03 Richard Biener <rguenther@suse.de>
6456 PR tree-optimization/87197
6457 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
6458 visited. CSE the VN_INFO hashtable lookup.
6460 PR tree-optimization/87169
6461 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
6462 iterating make sure there's no extra backedges from irreducible
6463 regions feeding the header. Mark the destination block
6466 2018-09-03 Martin Liska <mliska@suse.cz>
6469 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
6470 * common/common-targhooks.c (default_get_valid_option_values):
6472 * common/common-targhooks.h (default_get_valid_option_values):
6474 * common/config/i386/i386-common.c: Move processor_target_table
6476 (ix86_get_valid_option_values): New function.
6477 (TARGET_GET_VALID_OPTION_VALUES): New macro.
6478 * config/i386/i386.c (struct ptt): Move to i386-common.c.
6479 (PTA_*): Move all defined masks into i386-common.c.
6480 (ix86_function_specific_restore): Use new processor_cost_table.
6481 * config/i386/i386.h (struct ptt): Moved from i386.c.
6482 (struct pta): Likewise.
6483 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
6484 * doc/tm.texi.in: Likewise.
6485 * opt-suggestions.c (option_proposer::suggest_option):
6486 Pass prefix to build_option_suggestions.
6487 (option_proposer::get_completions): Likewise.
6488 (option_proposer::build_option_suggestions): Use the new target
6490 * opts.c (struct option_help_tuple): New struct.
6491 (print_filtered_help): Use the new target hook.
6493 2018-09-03 Martin Liska <mliska@suse.cz>
6496 * predict.c (set_even_probabilities): Add likely_edges
6497 argument and handle cases where we have precisely one
6499 (combine_predictions_for_bb): Catch also likely_edges.
6500 (tree_predict_by_opcode): Handle gswitch statements.
6501 * tree-cfg.h (find_case_label_for_value): New declaration.
6502 (find_taken_edge_switch_expr): Likewise.
6503 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
6504 Find pivot in decision tree based on probabily, not by number of
6507 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
6509 * doc/standards.texi (Standards): Update Objective-C reference.
6511 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
6513 * doc/install.texi (Prerequisites): Update link for MPC.
6515 2018-09-01 Michael Matz <matz@suse.de>
6517 PR tree-optimization/87074
6518 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
6519 PHIs for outer-loop uses.
6521 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
6523 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
6524 * doc/invoke.texi (C Dialect Options): Ditto.
6526 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
6528 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
6530 2018-08-31 Richard Biener <rguenther@suse.de>
6532 PR tree-optimization/87168
6533 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
6534 (rpo_elim::eliminate_avail): When OP was not visited it must
6537 2018-08-31 David Malcolm <dmalcolm@redhat.com>
6539 * tree-vrp.c (copy_value_range): Convert param "from" from
6540 "value_range *" to "const value_range *".
6541 (range_is_null): Likewise for param "vr".
6542 (range_int_cst_p): Likewise.
6543 (range_int_cst_singleton_p): Likewise.
6544 (symbolic_range_p): Likewise.
6545 (value_ranges_intersect_p): Likewise for both params.
6546 (value_range_nonnegative_p): Likewise for param "vr".
6547 (value_range_constant_singleton): Likewise.
6548 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
6549 (extract_range_into_wide_ints): Likewise for param "vr".
6550 (extract_range_from_multiplicative_op): Likewise for params "vr0"
6552 (vrp_can_optimize_bit_op): Likewise.
6553 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
6555 (extract_range_from_unary_expr): Likewise.
6556 (debug_value_range): Likewise for param "vr".
6557 (value_range::dump): Add "const" qualifier.
6558 (vrp_prop::check_array_ref): Convert local "vr" from
6559 "value_range *" to "const value_range *".
6560 (vrp_prop::check_mem_ref): Likewise.
6561 (vrp_prop::visit_stmt): Likewise for local "old_vr".
6562 (vrp_intersect_ranges_1): Likewise for param "vr_1".
6563 (vrp_intersect_ranges): Likewise.
6564 (simplify_stmt_for_jump_threading): Likewise for local "vr".
6565 (vrp_prop::vrp_finalize): Likewise.
6566 * tree-vrp.h (value_range::dump): Add "const" qualifier.
6567 (vrp_intersect_ranges): Add "const" qualifier to params as above.
6568 (extract_range_from_unary_expr): Likewise.
6569 (value_range_constant_singleton): Likewise.
6570 (symbolic_range_p): Likewise.
6571 (copy_value_range): Likewise.
6572 (extract_range_from_binary_expr_1): Likewise.
6573 (range_int_cst_p): Likewise.
6574 (vrp_set_zero_nonzero_bits): Likewise.
6575 (range_int_cst_singleton_p): Likewise.
6577 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
6579 * config/aarch64/arm_neon.h (vabsd_s64): New.
6580 (vnegd_s64): Likewise.
6582 2018-08-31 Martin Jambor <mjambor@suse.cz>
6584 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
6586 2018-08-31 Martin Liska <mliska@suse.cz>
6588 * ipa-icf.c (sem_item::add_type): Use
6589 sem_item::m_type_hash_cache.
6590 * ipa-icf.h: Move the cache from sem_item_optimizer
6593 2018-08-31 Nathan Sidwell <nathan@acm.org>
6595 * doc/extend.texi (Backwards Compatibility): Remove implicit
6596 extern C leeway of () being (...).
6598 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6600 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
6602 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
6606 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
6608 2018-08-31 Jakub Jelinek <jakub@redhat.com>
6611 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
6612 gen_int_mode. Formatting fixes.
6614 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
6616 * target.def (custom_function_descriptors): Improve documentation.
6617 * doc/tm.texi.in (Trampolines): Expand discussion of function
6618 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
6619 beginning of the section.
6620 * doc/tm.texi: Regenerated.
6622 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
6624 * cfg.h (class auto_edge_flag): Spell out the template-id of the
6625 base class in the initializer list. This is a workaround for
6626 building with older GCC.
6627 (class auto_bb_flag): Likewise.
6629 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
6631 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
6632 (altivec_vcmpequ<VI_char>_p): Remove star.
6633 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
6635 (expand_strncmp_vec_sequence): New function.
6636 (emit_final_str_compare_vec): New function.
6637 (expand_strn_compare): Add support for vector strncmp.
6638 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
6639 length specification to bytes.
6640 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
6641 (vcmpnezb_p): New pattern.
6642 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
6643 for option -mstring-compare-inline-limit.
6645 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
6647 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
6648 (PTA_SKYLAKE): Add PTA_AES.
6649 (PTA_GOLDMONT): Likewise.
6651 2018-08-29 Jan Hubicka <jh@suse.cz>
6654 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
6655 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
6657 2018-08-29 Jan Hubicka <jh@suse.cz>
6659 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
6661 (hash_tree): Do not visit TYPE_STUB_DECL.
6662 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
6663 stream TYPE_STUB_DECL.
6664 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
6665 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
6666 after free_lang_data.
6667 (type_in_anonymous_namespace_p): Likewise.
6669 2018-08-29 Jan Hubicka <jh@suse.cz>
6671 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
6672 comment that it has to be even number.
6673 (class sreal): Change m_sig type to int32_t.
6674 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
6675 int64_t for temporary calculations.
6676 (sreal_verify_basics): Drop one bit from minimum and maximum.
6678 2018-08-30 Richard Biener <rguenther@suse.de>
6680 PR tree-optimization/87147
6681 * tree-ssa-sccvn.c (SSA_VISITED): New function.
6682 (visit_phi): When the degenerate result is from the backedge and
6683 we didn't visit its definition yet drop to VARYING.
6684 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
6686 2018-08-29 Jan Hubicka <jh@suse.cz>
6688 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
6690 (hash_tree): Likewise.
6692 2018-08-29 Jan Hubicka <jh@suse.cz>
6694 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
6695 and TYPE_NEXT_REF_TO.
6697 2018-08-29 Jan Hubicka <jh@suse.cz>
6699 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
6700 comment that it has to be even number.
6701 (class sreal): Change m_sig type to int32_t.
6702 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
6703 int64_t for temporary calculations.
6704 (sreal_verify_basics): Drop one bit from minimum and maximum.
6706 2018-08-30 Tamar Christina <tamar.christina@arm.com>
6708 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
6710 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
6713 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
6714 if to_add is negative.
6716 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
6719 * builtins.c (c_strlen): Improve range checks.
6721 2018-08-29 Martin Sebor <msebor@redhat.com>
6722 Jeff Law <law@redhat.com>
6724 PR tree-optimization/86714
6725 PR tree-optimization/86711
6726 * builtins.c (c_strlen): Add arguments to call to string_constant.
6727 * expr.c (string_constant): Add argument. Detect missing nul
6728 terminator and outermost declaration it's missing in.
6729 * expr.h (string_constant): Add argument.
6730 * fold-const.c (read_from_constant_string): Add arguments to call to
6732 (c_getstr): Likewise.
6733 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
6735 * tree-ssa-strlen.c (get_stridx): Likewise.
6737 2018-08-29 Jan Hubicka <jh@suse.cz>
6739 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
6740 Do not stream DECL_VINDEX.
6741 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
6742 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
6743 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
6745 2018-08-29 Richard Biener <rguenther@suse.de>
6747 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
6748 virtual operands that are not default defs to honor region
6750 (rpo_vn_valueize): Remove ineffective code here.
6752 2018-08-29 Richard Biener <rguenther@suse.de>
6754 PR tree-optimization/87132
6755 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
6756 when skipping defs reachable over backedges.
6758 2018-08-29 Richard Biener <rguenther@suse.de>
6760 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
6761 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
6762 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
6763 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
6764 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
6765 (vn_reference_lookup_3): Remove use of const_parms.
6766 (free_rpo_vn): Do not free const_parms.
6767 (do_rpo_vn): Do not call init_const_parms.
6768 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
6769 SSA_NAME_POINTS_TO_READONLY_MEMORY.
6770 (call_may_clobber_ref_p_1): Likewise.
6772 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
6775 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
6777 (-ftree-scev-cprop): Document.
6779 2018-08-29 Jan Hubicka <jh@suse.cz>
6781 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
6783 (sreal constructor): Update.
6784 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
6785 sreal:operator/): Update.
6787 2018-08-29 Martin Liska <mliska@suse.cz>
6789 * tree-switch-conversion.c (switch_conversion::expand):
6790 Strenghten assumption about gswitch statements.
6792 2018-08-29 Richard Biener <rguenther@suse.de>
6794 PR tree-optimization/87117
6795 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
6796 re-value-number released SSA VDEFs.
6798 2018-08-29 Richard Biener <rguenther@suse.de>
6800 PR tree-optimization/87126
6801 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
6803 2018-08-28 Jim Wilson <jimw@sifive.com>
6805 * config/riscv/pic.md: Rewrite.
6806 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
6808 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
6809 (SOFTF, default_load, softload, softstore): New.
6811 2018-08-28 Jeff Law <law@redhat.com>
6813 * fold-const.c (fold_binary_loc): Remove recently added assert.
6815 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
6817 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
6818 to OP parmeter of generated function.
6820 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
6822 PR tree-optimization/87009
6823 * match.pd: Add boolean optimizations.
6825 2018-08-28 Martin Sebor <msebor@redhat.com>
6828 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
6829 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
6830 (pass_walloca::gate): Use it.
6831 (alloca_call_type): Same.
6832 (pass_walloca::execute): Same.
6833 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
6835 2018-08-28 David Malcolm <dmalcolm@redhat.com>
6837 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
6838 GCC_VERSION for usage of "__gcc_dump_printf__" format from
6841 2018-08-28 Richard Biener <rguenther@suse.de>
6843 PR tree-optimization/87124
6844 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
6845 constants before looking up avail.
6847 2018-08-28 Jakub Jelinek <jakub@redhat.com>
6850 * calls.c (maybe_warn_nonstring_arg): Punt early if
6851 warn_stringop_overflow is zero. Don't call get_range_strlen
6852 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
6853 Swap comparison operands to have constants on rhs. Only use
6854 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
6855 increment lenrng[0].
6857 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
6859 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
6860 use of tree_to_shwi. Remove duplicated test for the size being
6861 a whole number of bytes.
6863 2018-08-28 Richard Biener <rguenther@suse.de>
6865 PR tree-optimization/87117
6866 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
6867 Handle removed stmt without LHS (GIMPLE_NOP).
6869 2018-08-28 Richard Biener <rguenther@suse.de>
6871 PR tree-optimization/87117
6872 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
6873 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
6875 2018-08-28 Richard Biener <rguenther@suse.de>
6877 PR tree-optimization/87117
6878 * tree-ssa-pre.c (compute_avail): Do not make expressions
6879 with predicated values available.
6880 (get_expr_value_id): Assert we do not run into predicated value
6883 2018-08-28 Richard Biener <rguenther@suse.de>
6885 PR tree-optimization/87117
6886 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
6887 get virtual operands.
6888 (get_expr_operands): Handle STRING_CST like other decls.
6890 2018-08-28 Martin Liska <mliska@suse.cz>
6892 * tree.h: Update documentation of fndecl_built_in_p
6896 2018-08-27 Jeff Law <law@redhat.com>
6897 PR tree-optimization/87110
6898 * tree-ssa-dse.c (compute_trims): Handle non-constant
6901 2018-08-27 Martin Sebor <msebor@redhat.com>
6903 PR tree-optimization/86914
6904 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
6906 2018-08-27 Martin Sebor <msebor@redhat.com>
6908 PR tree-optimization/87112
6909 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
6910 the type of the bound argument.
6912 2018-08-27 Jeff Law <law@redhat.com>
6914 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
6915 type does not have a TYPE_SIZE_UNIT.
6917 2018-08-27 Steve Ellcey <sellcey@cavium.com>
6919 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
6920 with include of backend.h.
6922 2018-08-27 Richard Biener <rguenther@suse.de>
6924 PR tree-optimization/86927
6925 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
6926 use const cond reduction code.
6928 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
6930 PR tree-optimization/85758
6931 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
6933 2018-08-27 David Malcolm <dmalcolm@redhat.com>
6936 * diagnostic-show-locus.c (class layout_range): Update for
6937 conversion of show_caret_p to a tri-state.
6938 (layout_range::layout_range): Likewise.
6939 (make_range): Likewise.
6940 (layout::maybe_add_location_range): Likewise.
6941 (layout::should_print_annotation_line_p): Don't show annotation
6942 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
6943 (layout::get_state_at_point): Update for conversion of
6944 show_caret_p to a tri-state. Bail out early for
6945 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
6946 underlining or source colorization.
6947 (gcc_rich_location::add_location_if_nearby): Update for conversion
6948 of show_caret_p to a tri-state.
6949 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
6950 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
6952 (selftest::test_one_liner_labels): Likewise.
6953 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
6954 conversion of show_caret_p to a tri-state.
6955 * pretty-print.c (text_info::set_location): Likewise.
6956 * pretty-print.h (text_info::set_location): Likewise.
6957 * substring-locations.c (format_warning_n_va): Likewise.
6958 * tree-diagnostic.c (default_tree_printer): Likewise.
6959 * tree-pretty-print.c (newline_and_indent): Likewise.
6961 2018-08-27 David Malcolm <dmalcolm@redhat.com>
6964 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
6965 line above for line-insertion fix-it hints.
6966 (selftest::test_fixit_insert_containing_newline): Update the
6967 expected results, and add a test with line-numbering enabled.
6969 2018-08-27 Martin Liska <mliska@suse.cz>
6972 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
6973 params with DECL_HAS_VALUE_EXPR_P.
6975 2018-08-27 Martin Liska <mliska@suse.cz>
6977 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
6978 selected expansion strategy.
6980 2018-08-27 Martin Liska <mliska@suse.cz>
6982 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
6983 * builtins.c (is_builtin_fn): Likewise.
6984 * attribs.c (diag_attr_exclusions): Use new function
6985 fndecl_built_in_p and remove check for FUNCTION_DECL if
6987 (builtin_mathfn_code): Likewise.
6988 (fold_builtin_expect): Likewise.
6989 (fold_call_expr): Likewise.
6990 (fold_builtin_call_array): Likewise.
6991 (fold_call_stmt): Likewise.
6992 (set_builtin_user_assembler_name): Likewise.
6993 (is_simple_builtin): Likewise.
6994 * calls.c (gimple_alloca_call_p): Likewise.
6995 (maybe_warn_nonstring_arg): Likewise.
6996 * cfgexpand.c (expand_call_stmt): Likewise.
6997 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
6998 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
6999 (cgraph_node::verify_node): Likewise.
7000 * cgraphclones.c (build_function_decl_skip_args): Likewise.
7001 (cgraph_node::create_clone): Likewise.
7002 * config/arm/arm.c (arm_insert_attributes): Likewise.
7003 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
7004 * dse.c (scan_insn): Likewise.
7005 * expr.c (expand_expr_real_1): Likewise.
7006 * fold-const.c (operand_equal_p): Likewise.
7007 (fold_binary_loc): Likewise.
7008 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
7009 * gimple-low.c (lower_stmt): Likewise.
7010 * gimple-pretty-print.c (dump_gimple_call): Likewise.
7011 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7013 * gimple.c (gimple_build_call_from_tree): Likewise.
7014 (gimple_call_builtin_p): Likewise.
7015 (gimple_call_combined_fn): Likewise.
7016 * gimplify.c (gimplify_call_expr): Likewise.
7017 (gimple_boolify): Likewise.
7018 (gimplify_modify_expr): Likewise.
7019 (gimplify_addr_expr): Likewise.
7020 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
7021 * ipa-cp.c (determine_versionability): Likewise.
7022 * ipa-fnsummary.c (compute_fn_summary): Likewise.
7023 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
7024 * ipa-split.c (visit_bb): Likewise.
7025 (split_function): Likewise.
7026 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
7027 * lto-cgraph.c (input_node): Likewise.
7028 * lto-streamer-out.c (write_symbol): Likewise.
7029 * omp-low.c (setjmp_or_longjmp_p): Likewise.
7030 (lower_omp_1): Likewise.
7031 * predict.c (strip_predict_hints): Likewise.
7032 * print-tree.c (print_node): Likewise.
7033 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
7034 * trans-mem.c (is_tm_irrevocable): Likewise.
7035 (is_tm_load): Likewise.
7036 (is_tm_simple_load): Likewise.
7037 (is_tm_store): Likewise.
7038 (is_tm_simple_store): Likewise.
7039 (is_tm_abort): Likewise.
7040 (tm_region_init_1): Likewise.
7041 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
7042 * tree-cfg.c (verify_gimple_call): Likewise.
7043 (move_stmt_r): Likewise.
7044 (stmt_can_terminate_bb_p): Likewise.
7045 * tree-eh.c (lower_eh_constructs_2): Likewise.
7046 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
7047 * tree-inline.c (remap_gimple_stmt): Likewise.
7048 (copy_bb): Likewise.
7049 (estimate_num_insns): Likewise.
7050 (fold_marked_statements): Likewise.
7051 * tree-sra.c (scan_function): Likewise.
7052 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
7053 (optimize_stack_restore): Likewise.
7054 (pass_fold_builtins::execute): Likewise.
7055 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
7056 (mark_all_reaching_defs_necessary_1): Likewise.
7057 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
7058 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
7059 (pass_forwprop::execute): Likewise.
7060 * tree-ssa-loop-im.c (stmt_cost): Likewise.
7061 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
7062 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
7063 * tree-ssa-strlen.c (get_string_length): Likewise.
7064 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
7065 (find_func_aliases_for_call): Likewise.
7066 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
7067 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
7068 * tree-tailcall.c (find_tail_calls): Likewise.
7069 * tree.c (need_assembler_name_p): Likewise.
7070 (free_lang_data_in_decl): Likewise.
7071 (get_call_combined_fn): Likewise.
7072 * ubsan.c (is_ubsan_builtin_p): Likewise.
7073 * varasm.c (incorporeal_function_p): Likewise.
7074 * tree.h (DECL_BUILT_IN): Remove and replace with
7076 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
7077 (fndecl_built_in_p): New.
7079 2018-08-27 Martin Liska <mliska@suse.cz>
7081 PR tree-optimization/86847
7082 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
7083 Dump also subtree probability.
7084 (switch_decision_tree::do_jump_if_equal): New function.
7085 (switch_decision_tree::emit_case_nodes): Handle special
7086 situations in balanced tree that can be emitted much simpler.
7087 Fix calculation of probabilities that happen in tree expansion.
7088 * tree-switch-conversion.h (struct cluster): Add
7090 (struct simple_cluster): Likewise.
7091 (struct case_tree_node): Add new function has_child.
7092 (do_jump_if_equal): New.
7094 2018-08-27 Martin Liska <mliska@suse.cz>
7096 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
7097 Add new argument to bit_test_cluster constructor.
7098 (bit_test_cluster::emit): Set bits really number of values
7100 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
7101 probability argument.
7102 * tree-switch-conversion.h (struct bit_test_cluster):
7103 Add m_handles_entire_switch.
7105 2018-08-27 Martin Liska <mliska@suse.cz>
7107 PR tree-optimization/86702
7108 * tree-switch-conversion.c (jump_table_cluster::emit):
7109 Make probabilities even for values in jump table
7110 according to number of cases handled.
7111 (switch_decision_tree::compute_cases_per_edge): Pass
7112 argument to reset_out_edges_aux function.
7113 (switch_decision_tree::analyze_switch_statement): Likewise.
7114 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
7117 2018-08-27 Martin Liska <mliska@suse.cz>
7119 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
7120 cfun argument explicitly.
7121 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
7122 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
7123 function gimple_switch_default_bb.
7124 (convert_switch_statements):
7126 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
7127 * stmt.c (label_to_block_fn): Use label_to_block and pass
7128 cfun argument explicitly and use gimple_switch_label_bb.
7129 (expand_case): Likewise.
7130 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
7131 cfun argument explicitly. Likewise.
7132 (make_edges_bb): Likewise.
7133 (make_cond_expr_edges): Likewise.
7134 (get_cases_for_edge): Likewise.
7135 (make_gimple_switch_edges): Likewise.
7136 (label_to_block_fn): Likewise.
7137 (label_to_block): Likewise.
7138 (make_goto_expr_edges): Likewise.
7139 (make_gimple_asm_edges): Likewise.
7140 (main_block_label): Likewise.
7141 (group_case_labels_stmt): Likewise.
7142 (find_taken_edge_computed_goto): Likewise.
7143 (find_taken_edge_switch_expr): Likewise.
7144 (gimple_verify_flow_info): Likewise.
7145 (gimple_redirect_edge_and_branch): Likewise.
7146 (gimple_switch_label_bb): New function.
7147 (gimple_switch_default_bb): Likewise.
7148 (gimple_switch_edge): Likewise.
7149 (gimple_switch_default_edge): Likewise.
7150 * tree-cfg.h (label_to_block_fn): Remove and replace ...
7151 (label_to_block): ... with this.
7152 (gimple_switch_label_bb): New.
7153 (gimple_switch_default_bb): Likewise.
7154 (gimple_switch_edge): Likewise.
7155 (gimple_switch_default_edge): Likewise.
7156 * tree-cfgcleanup.c (convert_single_case_switch): Use
7157 new gimple functions and pass new argument to label_to_block.
7158 (cleanup_control_flow_bb):
7159 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
7160 cfun argument explicitly.
7161 (make_eh_edges): Likewise.
7162 (redirect_eh_dispatch_edge): Likewise.
7163 (lower_resx): Likewise.
7164 (lower_eh_dispatch): Likewise.
7165 (maybe_remove_unreachable_handlers): Likewise.
7166 (unsplit_eh): Likewise.
7167 (cleanup_empty_eh): Likewise.
7168 (verify_eh_edges): Likewise.
7169 (verify_eh_dispatch_edge): Likewise.
7170 * tree-ssa-dom.c (record_edge_info): Likewise.
7171 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
7172 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
7173 (thread_through_normal_block): Likewise.
7174 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
7175 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
7176 * tree-switch-conversion.c (switch_conversion::collect): Use new
7178 (switch_conversion::check_final_bb): Likewise.
7179 (switch_conversion::gather_default_values): Pass new argument
7181 (switch_conversion::build_constructors): Likewise.
7182 (switch_decision_tree::compute_cases_per_edge): Use new
7183 gimple_switch_edge function.
7184 (switch_decision_tree::analyze_switch_statement): Pass new argument
7186 (switch_decision_tree::try_switch_expansion): Use
7187 gimple_switch_default_edge.
7188 * tree-vrp.c (find_switch_asserts): Pass new argument
7190 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
7191 (vr_values::simplify_switch_using_ranges): Likewise.
7193 2018-08-27 Richard Biener <rguenther@suse.de>
7195 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
7196 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
7198 * tree-ssa-sccvn.h (struct vn_pval): New structure.
7199 (struct vn_nary_op_s): Add unwind_to member. Add
7200 predicated_values flag and put result into a union together
7201 with a linked list of vn_pval.
7202 (struct vn_ssa_aux): Add name member to make maintaining
7203 a map of SSA name to vn_ssa_aux possible. Remove no longer
7204 needed info, dfsnum, low, visited, on_sccstack, use_processed
7205 and range_info_anti_range_p members.
7206 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
7207 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
7209 (vn_valueize): New global.
7210 (vn_context_bb): Likewise.
7211 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
7212 VN_INFO_PTR_INFO): Remove.
7213 * tree-ssa-sccvn.c: ... (rewrite)
7214 (pass_fre::execute): For -O2+ initialize loops and run
7215 RPO VN in optimistic mode (iterating). For -O1 and -Og
7216 run RPO VN in non-optimistic mode.
7217 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
7218 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
7219 * doc/invoke.texi (sccvn-max-scc-size): Remove.
7220 (rpo-vn-max-loop-depth): Document.
7221 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
7222 when valuezing the VUSE signals we walked out of the region.
7223 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
7224 (phi_translate): Set VN context block to use for availability
7226 (compute_avail): Likewise.
7227 (pre_valueize): New function.
7228 (pass_pre::execute): Adjust to the RPO VN API.
7230 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
7231 (propagate_constants_for_unrolling): Remove.
7232 (tree_unroll_loops_completely): Perform value-numbering
7233 on the unrolled bodies loop parent.
7235 2018-08-27 Richard Biener <rguenther@suse.de>
7237 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
7238 for partial antic compute.
7240 2018-08-27 Jakub Jelinek <jakub@redhat.com>
7242 PR rtl-optimization/87065
7243 * combine.c (simplify_if_then_else): Formatting fix.
7244 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
7246 (known_cond): Don't return const_true_rtx for vector modes. Use
7247 CONST0_RTX instead of const0_rtx. Formatting fixes.
7249 2018-08-27 Martin Liska <mliska@suse.cz>
7251 PR gcov-profile/87069
7252 * gcov.c (process_file): Record files already processed
7253 and warn about a file being processed multiple times.
7255 2018-08-27 Martin Liska <mliska@suse.cz>
7258 * config/aarch64/aarch64.c (aarch64_override_options_internal):
7259 Set default values for x_aarch64_*_string strings.
7260 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
7261 prefix. For -mabi do not print '=ABI' in help and use
7262 <option_value> format for -msve-vector-bits and -moverride
7265 2018-08-26 Jeff Law <law@redhat.com>
7267 * config/mips/frame-header-opt.c: Include "backend.h" rather than
7270 2018-08-26 Marek Polacek <polacek@redhat.com>
7272 PR c++/87029, Implement -Wredundant-move.
7273 * doc/invoke.texi: Document -Wredundant-move.
7275 2018-08-25 Martin Sebor <msebor@redhat.com>
7277 PR tree-optimization/87059
7278 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
7279 to the same type as the other.
7280 * fold-const.c (fold_binary_loc): Assert expectation.
7282 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
7284 * config/darwin.c (machopic_legitimize_pic_address): Clean up
7285 extraneous parentheses, dead code section and formatting.
7287 2018-08-24 David Malcolm <dmalcolm@redhat.com>
7290 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
7291 wide enough for jumps in the line-numbering to be visible.
7292 (layout::print_gap_in_line_numbering): New member function.
7293 (layout::calculate_line_spans): When using line numbering, merge
7294 line spans that are only 1 line apart.
7295 (diagnostic_show_locus): When printing line numbers, show gaps in
7296 line numbering directly, rather than printing headers.
7297 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
7298 line-numbering with multiple line spans.
7299 (selftest::test_fixit_insert_containing_newline_2): Add test of
7300 line-numbering, in which the spans are close enough to be merged.
7302 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
7304 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
7305 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
7306 * tree-vrp.c (range_is_nonnull): Remove.
7307 (range_includes_zero_p): Accept value_range instead of min/max.
7308 (extract_range_from_binary_expr_1): Do not early bail on
7310 Use range_includes_zero_p instead of range_is_nonnull.
7311 (extract_range_from_unary_expr): Use range_includes_zero_p instead
7312 of range_is_nonnull.
7313 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
7314 special case VR_ANTI_RANGE.
7315 (vrp_finalize): Same.
7316 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
7318 (range_is_nonnull): Remove.
7319 * vr-values.c (vrp_stmt_computes_nonzero): Use
7320 range_includes_zero_p instead of range_is_nonnull.
7321 (extract_range_basic): Pass value_range to range_includes_zero_p
7322 instead of range_is_nonnull.
7324 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
7326 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
7327 * emit-rtl.h (rtl_data): Remove return_bnd.
7328 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
7329 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
7330 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
7331 (POINTER_BOUNDS_MODE): Remove definition.
7332 (make_pointer_bounds_mode): Remove.
7333 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
7334 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
7335 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
7336 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
7337 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
7338 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
7339 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
7341 * config/i386/i386-modes.def (BND32, BND64): Remove.
7342 * config/i386/i386.c (dbx_register_map): Remove bound registers.
7343 (dbx64_register_map): Ditto.
7344 (svr4_dbx_register_map): Ditto.
7345 (indirect_thunk_bnd_needed): Remove.
7346 (indirect_thunks_bnd_used): Ditto.
7347 (indirect_return_bnd_needed): Ditto.
7348 (indirect_return_via_cx_bnd): Ditto.
7349 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
7350 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
7351 (output_indirect_thunk): Ditto. Remove need_prefix argument.
7352 (output_indirect_thunk_function): Remove handling of
7353 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
7354 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
7355 (ix86_save_reg): Remove handling of crtl->return_bnd.
7356 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
7357 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
7358 and UNSPEC_BNDLX_ADDR.
7359 (ix86_output_indirect_branch_via_reg): Remove handling of
7360 indirect_thunk_prefix_bnd.
7361 (ix86_output_indirect_branch_via_push): Ditto.
7362 (ix86_output_function_return): Ditto.
7363 (ix86_output_indirect_function_return): Ditto.
7364 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
7365 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
7366 (CALL_USED_REGISTERS): Ditto.
7367 (REG_ALLOC_ORDER): Update for removal of bound registers.
7368 (HI_REGISTER_NAMES): Ditto.
7369 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
7370 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
7371 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
7372 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
7373 (FIRST_PSEUDO_REG): Update.
7374 (BND): Remove mode iterator.
7375 * config/i386/predicates.md (bnd_mem_operator): Remove.
7377 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
7379 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
7382 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
7384 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
7385 the case in which the permute needs only a single element and
7386 repeats for every vector of the result. Extend that case to
7387 handle variable-length vectors.
7388 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
7390 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
7393 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
7394 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
7396 2018-08-24 Richard Biener <rguenther@suse.de>
7398 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
7399 bb_flags_allocated members.
7400 (auto_flag): New RAII class for allocating flags.
7401 (auto_edge_flag): New RAII class for allocating edge flags.
7402 (auto_bb_flag): New RAII class for allocating bb flags.
7403 * cfgloop.c (verify_loop_structure): Allocate temporary edge
7405 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
7406 in favor of temporarily allocated BB flag.
7407 * hsa-brig.c: Re-order includes.
7408 * hsa-dump.c: Likewise.
7409 * hsa-regalloc.c: Likewise.
7410 * print-rtl.c: Likewise.
7411 * profile-count.c: Likewise.
7413 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
7416 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
7419 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
7422 * wide-int-range.cc (wide_int_range_div): Do not ignore result
7423 from wide_int_range_multiplicative_op.
7425 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7427 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
7430 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
7432 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
7435 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
7437 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
7438 full GENERIC expression used for replacement.
7440 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
7442 * tree-vrp.c (abs_extent_range): Remove.
7443 (extract_range_into_wide_ints): Pass wide ints by reference.
7444 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
7445 Pass wide ints by reference in all calls to
7446 extract_range_into_wide_ints.
7447 * wide-int-range.cc (wide_int_range_div): New.
7448 * wide-int-range.h (wide_int_range_div): New.
7449 (wide_int_range_includes_zero_p): New.
7450 (wide_int_range_zero_p): New.
7452 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
7454 * config/aarch64/aarch64.md (arches): New enum.
7455 (arch): New enum attr.
7456 (arch_enabled): New attr.
7457 (enabled): Now uses arch_enabled only.
7458 (simd, sve, fp16): Removed attribute.
7459 (fp): Attr now defined in terms of 'arch'.
7460 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
7461 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
7462 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
7463 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
7464 attributes into 'arch'.
7465 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
7466 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
7467 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
7468 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
7469 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
7470 'simd' attribute into 'arch'.
7471 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
7472 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
7473 Convert use of 'fp' attribute to 'arch'.
7474 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
7475 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
7477 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
7478 (different modes) Merge 'fp' and 'simd' into 'arch'.
7479 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
7482 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
7484 PR rtl-optimization/87026
7485 * expmed.c (canonicalize_comparison): If we can no longer create
7486 pseudoregisters, don't.
7488 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
7491 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
7493 * config/arm/arm.c (speculation_barrier_libfunc): New static
7495 (arm_init_libfuncs): Initialize it.
7496 (arm_emit_speculation_barrier): New function.
7497 * config/arm/arm.md (speculation_barrier): Call
7498 arm_emit_speculation_barrier for architectures that do not have
7500 (speculation_barrier_insn): Only match on Armv7 or later.
7502 2018-08-23 Richard Biener <rguenther@suse.de>
7505 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
7508 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
7510 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
7511 of single-vector TBLs.
7512 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
7515 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
7518 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
7519 aarch64_evpc_tbl guard.
7521 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
7523 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
7526 2018-08-22 Martin Sebor <msebor@redhat.com>
7529 * tree-pretty-print.c (pretty_print_string): Add argument.
7530 (dump_generic_node): Call to pretty_print_string with string size.
7532 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
7534 PR rtl-optimization/86771
7535 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
7536 of two SETs into those two SETs, one to be placed at i2, if that SETs
7537 destination is modified between i2 and i3.
7539 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
7541 PR tree-optimization/86725
7542 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
7544 (vect_analyze_scalar_cycles_1): Check it.
7546 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
7548 PR tree-optimization/86725
7549 * tree-vect-loop.c (vect_is_simple_reduction): When treating
7550 an outer loop phi as a double reduction, make sure that the
7551 single user of the phi result is an inner loop phi.
7553 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
7555 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
7556 grouped stores with gaps to a strided group.
7558 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
7560 * tree-vect-stmts.c (get_group_load_store_type)
7561 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
7562 first statement in a group.
7564 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
7566 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
7567 the sequence used in gcc/gcc.c.
7569 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
7572 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
7575 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
7577 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
7578 Darwin10-specific unwinder-shim.
7579 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
7580 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
7581 New to cater for Darwin10 Rosetta.
7583 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
7585 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
7586 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
7588 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
7593 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
7594 (dwarf2out_switch_text_section): Generate a local label for the
7595 second function sub-section and apply it as the second FDE start
7597 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
7598 second sub-section start.
7600 2018-08-22 Richard Biener <rguenther@suse.de>
7602 PR tree-optimization/86988
7603 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
7605 2018-08-22 Richard Biener <rguenther@suse.de>
7607 PR tree-optimization/86945
7608 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
7610 2018-08-22 Alexandre Oliva <oliva@adacore.com>
7612 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
7613 a comment about how uses of r2 for .sdata2 come about.
7615 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
7617 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
7619 2018-08-21 Marek Polacek <polacek@redhat.com>
7621 PR c++/86981, Implement -Wpessimizing-move.
7622 * doc/invoke.texi: Document -Wpessimizing-move.
7624 2018-08-21 Jan Hubicka <jh@suse.cz>
7626 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
7627 * tree.h (is_redundant_typedef): Remove.
7628 * dwarf2out.c (is_redundant_typedef): Turn into static function.
7630 2018-08-21 Jan Hubicka <jh@suse.cz>
7632 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
7635 2018-08-21 Tamar Christina <tamar.christina@arm.com>
7637 * expmed.c (extract_low_bits): Reject invalid subregs early.
7639 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
7642 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
7645 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
7647 * config/vxworks.h: Guard vxworks_asm_out_constructor and
7648 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
7649 * config/vxworks.c: Likewise.
7651 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
7653 * config/vxworks.c: Set targetm.have_ctors_dtors
7654 if HAVE_INITFINI_ARRAY_SUPPORT.
7655 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
7656 if HAVE_INITFINI_ARRAY_SUPPORT.
7658 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
7660 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
7661 default search path for VxWorks < 7.
7663 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
7665 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
7666 (get_format_string): Refer to c_getstr.
7668 2018-08-21 Tom de Vries <tdevries@suse.de>
7670 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
7671 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
7672 (debuginfo_early_stop): Declare.
7673 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
7674 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
7675 (debuginfo_early_stop): New function.
7676 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
7677 and debuginfo_early_stop.
7678 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
7679 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
7680 (general_init): Call debuginfo_early_init.
7681 (finalize): Call debuginfo_fini.
7682 (do_compile): Call debuginfo_init.
7683 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
7685 (@item -fdump-debug, @item -fdump-earlydebug): Add.
7687 2018-08-21 Tom de Vries <tdevries@suse.de>
7689 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
7690 flag_dump_noaddr and flag_dump_unnumbered.
7692 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
7694 * wide-int-range.cc (wide_int_range_abs): New.
7695 (wide_int_range_order_set): Rename from wide_int_range_min_max.
7696 * wide-int-range.h (wide_int_range_abs): New.
7697 (wide_int_range_min_max): New.
7698 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
7699 case to call wide_int_range_abs.
7700 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
7701 (extract_range_from_abs_expr): Delete.
7703 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
7706 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
7707 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
7710 2018-08-20 Nathan Sidwell <nathan@acm.org>
7711 Jeff Law <law@redhat.com>
7713 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
7714 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
7716 2018-08-20 David Malcolm <dmalcolm@redhat.com>
7719 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
7720 (decl_attributes): Likewise.
7721 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
7723 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
7724 * diagnostic-core.h (class auto_diagnostic_group): New class.
7725 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
7726 (diagnostic_report_diagnostic): Handle the first diagnostics within
7728 (emit_diagnostic): Add auto_diagnostic_group instance.
7730 (inform_n): Likewise.
7731 (warning): Likewise.
7732 (warning_at): Likewise.
7733 (warning_n): Likewise.
7734 (pedwarn): Likewise.
7735 (permerror): Likewise.
7737 (error_n): Likewise.
7738 (error_at): Likewise.
7740 (fatal_error): Likewise.
7741 (internal_error): Likewise.
7742 (internal_error_no_backtrace): Likewise.
7743 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
7744 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
7745 * diagnostic.h (struct diagnostic_context): Add fields
7746 "diagnostic_group_nesting_depth",
7747 "diagnostic_group_emission_count", "begin_group_cb",
7749 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
7750 Add auto_diagnostic_group instance(s).
7751 (find_explicit_erroneous_behavior): Likewise.
7752 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
7753 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
7754 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
7755 (gimplify_va_arg_expr): Likewise.
7756 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
7757 (HSA_SORRY_AT): Likewise.
7758 * ipa-devirt.c (compare_virtual_tables): Likewise.
7759 (warn_odr): Likewise.
7760 * multiple_target.c (expand_target_clones): Likewise.
7761 * opts-common.c (cmdline_handle_error): Likewise.
7762 * reginfo.c (globalize_reg): Likewise.
7763 * substring-locations.c (format_warning_n_va): Likewise.
7764 * tree-inline.c (expand_call_inline): Likewise.
7765 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
7766 * tree-ssa-loop-niter.c
7767 (do_warn_aggressive_loop_optimizations): Likewise.
7768 * tree-ssa-uninit.c (warn_uninit): Likewise.
7769 * tree.c (warn_deprecated_use): Likewise.
7771 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
7774 * config/i386/i386.md (eh_return): Always update EH return
7775 address in word_mode.
7777 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
7779 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
7780 TARGET_SPLIT_COMPLEX_ARG is defined.
7782 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
7784 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
7786 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
7789 * expr.c (expand_assignment): Assert that bitpos is positive.
7790 (store_field): Likewise
7791 (expand_expr_real_1): Make sure that bitpos is positive.
7792 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
7795 2018-08-20 Nathan Sidwell <nathan@acm.org>
7797 * Makefile.in (CPP_ID_DATA_H): Delete.
7798 (CPP_INTERNAL_H): Don't add it.
7799 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
7800 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
7802 2018-08-20 Richard Biener <rguenther@suse.de>
7804 PR tree-optimization/78655
7805 * tree-vrp.c (extract_range_from_binary_expr_1): Make
7806 pointer + offset nonnull if either operand is nonnull work.
7808 2018-08-20 Tom de Vries <tdevries@suse.de>
7810 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
7811 unless the referenced die describes the added property using
7812 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
7813 Otherwise, add a DW_AT_location to the referenced die.
7815 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
7818 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
7819 register_operand when calling ix86_set_reg_reg_cost.
7820 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
7821 Set *total to 0 for operands that satisfy x86_64_immediate_operand
7822 predicate and to 1 otherwise.
7824 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
7826 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
7827 emit a diagnostic that it is not supported and reset the option.
7828 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
7829 supported and consume the option. (ASM_FINAL_SPEC): New.
7831 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
7833 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
7836 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
7838 C-SKY port: Documentation
7840 * doc/extend.texi (C-SKY Function Attributes): New section.
7841 * doc/invoke.texi (Option Summary): Add C-SKY options.
7842 (C-SKY Options): New section.
7843 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
7845 2018-08-17 Jojo <jijie_rong@c-sky.com>
7846 Huibin Wang <huibin_wang@c-sky.com>
7847 Sandra Loosemore <sandra@codesourcery.com>
7848 Chung-Lin Tang <cltang@codesourcery.com>
7850 C-SKY port: Backend implementation
7852 * config/csky/*: New.
7853 * common/config/csky/*: New.
7855 2018-08-17 Jojo <jijie_rong@c-sky.com>
7856 Huibin Wang <huibin_wang@c-sky.com>
7857 Sandra Loosemore <sandra@codesourcery.com>
7858 Chung-Lin Tang <cltang@codesourcery.com>
7859 Andrew Jenner <andrew@codesourcery.com>
7861 C-SKY port: Configury
7863 * config.gcc (csky-*-*): New.
7864 * configure.ac: Add csky to targets for dwarf2 debug_line support.
7865 * configure: Regenerated.
7867 2018-08-17 David Malcolm <dmalcolm@redhat.com>
7869 * dump-context.h: Include "dumpfile.h".
7870 (dump_context::dump_printf_va): Convert final param from va_list
7871 to va_list *. Convert from ATTRIBUTE_PRINTF to
7872 ATTRIBUTE_GCC_DUMP_PRINTF.
7873 (dump_context::dump_printf_loc_va): Likewise.
7874 * dumpfile.c: Include "stringpool.h".
7875 (make_item_for_dump_printf_va): Delete.
7876 (make_item_for_dump_printf): Delete.
7877 (class dump_pretty_printer): New class.
7878 (dump_pretty_printer::dump_pretty_printer): New ctor.
7879 (dump_pretty_printer::emit_items): New member function.
7880 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
7882 (dump_pretty_printer::emit_item): New member function.
7883 (dump_pretty_printer::stash_item): New member function.
7884 (dump_pretty_printer::format_decoder_cb): New member function.
7885 (dump_pretty_printer::decode_format): New member function.
7886 (dump_context::dump_printf_va): Reimplement in terms of
7887 dump_pretty_printer.
7888 (dump_context::dump_printf_loc_va): Convert final param from va_list
7890 (dump_context::begin_scope): Reimplement call to
7891 make_item_for_dump_printf.
7892 (dump_printf): Update for change to dump_printf_va.
7893 (dump_printf_loc): Likewise.
7894 (selftest::test_capture_of_dump_calls): Convert "stmt" from
7895 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
7896 with %T, %E, and %G.
7897 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
7898 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
7899 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
7900 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
7901 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
7902 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
7903 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
7904 within a dump_printf_loc call to "%wu".
7905 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
7906 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
7907 missing space after "=".
7908 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
7909 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
7910 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
7911 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
7912 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
7913 duplicate "vectorized" from message.
7915 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
7917 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
7918 polyNxK_t element's TYPE_STRING_FLAG.
7920 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
7922 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
7923 (they were unnamed before). Fix comments.
7925 2018-08-17 Nathan Sidwell <nathan@acm.org>
7927 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
7929 2018-08-17 Richard Biener <rguenther@suse.de>
7931 PR tree-optimization/86841
7932 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
7934 2018-08-17 Martin Liska <mliska@suse.cz>
7936 * common.opt: Remove Warn, Init and Report for options with
7937 Ignore/Deprecated flag. Warning is done automatically for
7939 * config/i386/i386.opt: Likewise.
7940 * config/ia64/ia64.opt: Likewise.
7941 * config/rs6000/rs6000.opt: Likewise.
7942 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
7943 Remove usage of flag_check_pointer_bounds.
7944 * lto-wrapper.c (merge_and_complain): Do not handle
7945 OPT_fcheck_pointer_bounds.
7946 (append_compiler_options): Likewise.
7947 * opt-functions.awk: Do not handle Deprecated.
7948 * optc-gen.awk: Check that Var, Report and Init are not
7949 used for an option with Ignore/Deprecated flag.
7950 * opts-common.c (decode_cmdline_option): Do not report
7952 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
7954 * opts.h (struct cl_option): Remove cl_deprecated flag.
7955 (CL_ERR_DEPRECATED): Remove error enum value.
7957 2018-08-17 Richard Biener <rguenther@suse.de>
7960 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
7961 across a va-arg-pack using call adjust its return value accordingly.
7963 2018-08-16 Martin Sebor <msebor@redhat.com>
7965 PR tree-optimization/86853
7966 * gimple-ssa-sprintf.c (struct format_result): Rename member.
7967 (struct fmtresult): Add member and initialize it in ctors.
7968 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
7969 (format_string): Handle %S the same as %ls. Set MAYFAIL.
7970 (format_directive): Set POSUNDER4K when MAYFAIL is set.
7971 (parse_directive): Handle %C same as %c.
7972 (sprintf_dom_walker::compute_format_length): Adjust.
7973 (is_call_safe): Adjust.
7975 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
7977 * builtins.c (c_strlen): Add new parameter eltsize. Use it
7978 for determining how to count the elements.
7979 * builtins.h (c_strlen): Adjust prototype.
7980 * expr.c (string_constant): Add new parameter mem_size.
7981 Set *mem_size appropriately.
7982 * expr.h (string_constant): Adjust protoype.
7983 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
7984 * gimple-fold.h (get_range_strlen): Adjust prototype.
7985 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
7986 (format_string): Call get_string_length with eltsize.
7988 2018-08-16 David Malcolm <dmalcolm@redhat.com>
7990 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
7991 to emit the span, rather than setting it as the prefix.
7993 2018-08-16 David Malcolm <dmalcolm@redhat.com>
7995 * diagnostic-show-locus.c (layout::start_annotation_line): Add
7996 "margin_char" parameter, defaulting to space. Use it in place
7997 of pp_space for the initial part of the margin.
7998 (layout::print_leading_fixits): Use '+' when filling the margin
7999 of line-insertion fix-it hints.
8001 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
8003 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
8006 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
8008 * config/rs6000/altivec.md: Don't set length attribute to the default
8010 * config/rs6000/darwin.md: Ditto.
8011 * config/rs6000/dfp.md: Ditto.
8012 * config/rs6000/htm.md: Ditto.
8013 * config/rs6000/rs6000.md: Ditto.
8014 * config/rs6000/sync.md: Ditto.
8015 * config/rs6000/vsx.md: Ditto.
8017 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
8019 * config/rs6000/altivec.md: Don't set length attribute to the default
8020 value, for branch instructions.
8021 * config/rs6000/darwin.md: Ditto.
8022 * config/rs6000/rs6000.md: Ditto.
8024 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
8026 * config/rs6000/rs6000.md (length): Always define as const_int 4.
8027 (unnamed conditional branch define_insn): Set length to 4 or 8
8028 depending on offset.
8029 (<bd>_<mode>): Similar, for alternative 0.
8030 (<bd>tf_<mode>): Ditto.
8032 2018-08-16 Tamar Christina <tamar.christina@arm.com>
8034 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
8036 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
8038 * doc/rtl.texi: Replace old RTX class names with new names.
8041 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
8043 * expmed.h (canonicalize_comparison): New declaration.
8044 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
8045 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
8046 * optabs.c (prepare_cmp_insn): Likewise.
8047 * rtl.h (unsigned_condition_p): New function which checks if a
8048 comparison operator is unsigned.
8050 2018-08-16 Nathan Sidwell <nathan@acm.org>
8052 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
8053 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
8055 2018-08-16 Tamar Christina <tamar.christina@arm.com>
8058 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
8059 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
8060 (mov<mov>): ..this and enable unconditionally.
8062 2018-08-16 Tamar Christina <tamar.christina@arm.com>
8064 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
8066 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
8068 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
8069 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
8070 "Common" with "Target".
8072 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
8074 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
8075 * doc/invoke.texi (mmitigate-rop): Remove.
8076 * config/i386/i386.c: Do not include "regrename.h".
8077 (ix86_rop_should_change_byte_p, reg_encoded_number)
8078 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
8080 (ix86_reorg): Remove call to ix86_mitigate_rop.
8081 * config/i386/i386.md (attr "modrm_class"): Remove.
8082 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
8083 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
8084 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
8086 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
8088 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
8089 allow folding of mergeh() and mergel() for the float and double types.
8090 (fold_mergehl_helper): Rework to handle building a permute tree
8093 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
8095 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
8098 2018-08-15 David Malcolm <dmalcolm@redhat.com>
8100 * common.opt (fdiagnostics-show-labels): New option.
8101 * diagnostic-show-locus.c (class layout_range): Add field
8103 (class layout): Add field "m_show_labels_p".
8104 (layout_range::layout_range): Add param "label" and use it to
8106 (make_range): Pass in NULL for new "label" param of layout_range's
8108 (layout::layout): Initialize m_show_labels_p.
8109 (layout::maybe_add_location_range): Pass in loc_range->m_label
8110 when constructing layout_range instances.
8111 (struct line_label): New struct.
8112 (layout::print_any_labels): New member function.
8113 (layout::print_line): Call it if label-printing is enabled.
8114 (selftest::test_one_liner_labels): New test.
8115 (selftest::test_diagnostic_show_locus_one_liner): Call it.
8116 * diagnostic.c (diagnostic_initialize): Initialize
8117 context->show_labels_p.
8118 * diagnostic.h (struct diagnostic_context): Add field
8120 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
8121 -fno-diagnostics-show-labels.
8122 * dwarf2out.c (gen_producer_string): Add
8123 OPT_fdiagnostics_show_labels to the ignored options.
8124 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
8126 (gcc_rich_location::maybe_add_expr): Likewise.
8127 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
8128 label" param, defaulting to NULL.
8129 (gcc_rich_location::add_expr): Add "label" param.
8130 (gcc_rich_location::maybe_add_expr): Likewise.
8131 (class text_range_label): New class.
8132 (class range_label_for_type_mismatch): New class.
8133 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
8134 of format_warning_va.
8135 (fmtwarn_n): Likewise for new params of format_warning_n_va.
8136 * lto-wrapper.c (merge_and_complain): Add
8137 OPT_fdiagnostics_show_labels to the "pick one setting" options.
8138 (append_compiler_options): Likewise to the dropped options.
8139 (append_diag_options): Likewise to the passed-on options.
8140 * opts.c (common_handle_option): Handle the new option.
8141 * selftest-diagnostic.c
8142 (test_diagnostic_context::test_diagnostic_context): Enable
8144 * substring-locations.c: Include "gcc-rich-location.h".
8145 (format_warning_n_va): Add "fmt_label" and "param_label" params
8146 and use them as appropriate.
8147 (format_warning_va): Add "fmt_label" and "param_label" params,
8148 passing them on to format_warning_n_va.
8149 (format_warning_at_substring): Likewise.
8150 (format_warning_at_substring_n): Likewise.
8151 * substring-locations.h (format_warning_va): Add "fmt_label" and
8152 "param_label" params.
8153 (format_warning_n_va): Likewise.
8154 (format_warning_at_substring): Likewise.
8155 (format_warning_at_substring_n): Likewise.
8156 * toplev.c (general_init): Initialize global_dc->show_labels_p.
8158 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
8161 * builtins.c (expand_builtin_memcmp): Do not expand the call
8162 when overflow is detected.
8164 2018-08-15 Martin Sebor <msebor@redhat.com>
8166 PR tree-optimization/71625
8167 * config/aarch64/aarch64-builtins.c
8168 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
8170 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
8172 * config/s390/s390.c (s390_reorg): Remove loop.
8174 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
8177 (darwin_function_switched_text_sections): Delete.
8178 * gcc/config/darwin.h
8179 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
8181 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
8184 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
8185 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
8186 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
8188 2018-08-15 Martin Liska <mliska@suse.cz>
8190 PR tree-optimization/86925
8191 * predict.c (expr_expected_value_1): When taking
8192 later predictor, assign also probability.
8193 Use fold_build2_initializer_loc in order to fold
8194 the expression in -frounding-math.
8196 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
8198 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
8200 (expand_vec_perm_1): Try the new method.
8202 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
8205 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
8206 Check whether lra_live_max_point is 0 before dividing.
8208 2018-08-14 Martin Sebor <msebor@redhat.com>
8210 PR tree-optimization/86650
8211 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
8212 (vrp_prop::check_mem_ref): Same.
8214 2018-08-13 Liu Hao <lh_mouse@126.com>
8216 * pretty-print.c (eat_esc_sequence): Swap the foreground and
8217 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
8218 and clear it thereafter, as it only works for DBCS.
8220 2018-08-13 Liu Hao <lh_mouse@126.com>
8222 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
8223 handle returned by _get_osf_handle().
8225 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
8227 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
8228 for folding vec_perm.
8230 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
8232 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
8233 Add support for gimple-folding of vec_pack() and vec_unpack()
8236 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
8238 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
8239 vec_xst variants to the list.
8240 (rs6000_gimple_fold_builtin): Add support for folding unaligned
8241 vector loads and stores.
8243 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
8245 * config.gcc (rs6000-ibm-aix4.x): Delete.
8246 (rs6000-ibm-aix5.1): Delete.
8247 (rs6000-ibm-aix5.2): Delete.
8248 (rs6000-ibm-aix5.3): Delete.
8249 * config/rs6000/aix43.h: Delete.
8250 * config/rs6000/aix51.h: Delete.
8251 * config/rs6000/aix52.h: Delete.
8252 * config/rs6000/t-aix43: Delete.
8254 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
8256 * config/s390/s390.c (s390_decompose_constant_pool_ref):
8258 (s390_decompose_address): Factor out constant pool ref
8261 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
8263 * config/nds32/nds32-predicates.c
8264 (nds32_can_use_bclr_p): Change return type as bool.
8265 (nds32_can_use_bset_p): Ditto.
8266 (nds32_can_use_btgl_p): Ditto.
8267 (nds32_can_use_bitci_p): Ditto.
8268 * config/nds32/nds32-protos.h
8269 (nds32_can_use_bclr_p): Change declaration.
8270 (nds32_can_use_bset_p): Ditto.
8271 (nds32_can_use_btgl_p): Ditto.
8272 (nds32_can_use_bitci_p): Ditto.
8274 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
8276 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
8277 Support -msched-prolog-epilog option.
8278 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
8280 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
8282 * common/config/nds32/nds32-common.c
8283 (nds32_option_optimization_table): Enalbe -malways-align.
8285 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
8287 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
8289 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
8290 OPT_misr_secure_ case.
8291 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
8292 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
8293 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
8295 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
8296 (nds32_isr_info): New field security_level.
8297 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
8298 * config/nds32/nds32.md (return_internal): Consider critical attribute.
8299 * config/nds32/nds32.opt (misr-secure): New option.
8300 * config/nds32/nds32_init.inc: New file.
8301 * config/nds32/nds32_isr.h: New file.
8303 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
8305 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
8306 Update comment for atomic instructions.
8307 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
8308 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
8310 (atomic_loaddi): Revise fence expansion to only emit fence prior to
8311 load for __ATOMIC_SEQ_CST model.
8312 (atomic_loaddi_1): Remove float register target.
8313 (atomic_storedi): Handle CONST_INT values.
8314 (atomic_storedi_1): Remove float register source. Add special case
8316 (memory_barrier): New expander and insn.
8318 2018-08-11 Jakub Jelinek <jakub@redhat.com>
8320 PR tree-optimization/86835
8321 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
8322 new_stmt after def_gsi, make sure to insert new_square_stmt after
8323 that stmt, not 2 stmts before it.
8325 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
8328 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
8331 2018-08-10 Martin Liska <mliska@suse.cz>
8334 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
8336 * builtins.c (expand_builtin_expect_with_probability):
8338 (expand_builtin_expect_with_probability): New function.
8339 (build_builtin_expect_predicate): Add new argumnet probability
8340 for BUILT_IN_EXPECT_WITH_PROBABILITY.
8341 (fold_builtin_expect):
8344 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
8345 * builtins.h (fold_builtin_expect): Set new argument.
8346 * doc/extend.texi: Document __builtin_expect_with_probability.
8347 * doc/invoke.texi: Likewise.
8348 * gimple-fold.c (gimple_fold_call): Pass new argument.
8349 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
8350 also BUILT_IN_EXPECT_WITH_PROBABILITY.
8351 * predict.c (get_predictor_value): New function.
8352 (expr_expected_value): Add new argument probability. Assume
8353 that predictor and probability are always non-null.
8354 (expr_expected_value_1): Likewise. For __builtin_expect and
8355 __builtin_expect_with_probability set probability. Handle
8356 combination in binary expressions.
8357 (tree_predict_by_opcode): Simplify code by simply calling
8358 get_predictor_value.
8359 (pass_strip_predict_hints::execute): Add handling of
8360 BUILT_IN_EXPECT_WITH_PROBABILITY.
8361 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
8363 * tree.h (DECL_BUILT_IN_P): New function.
8365 2018-08-10 Martin Liska <mliska@suse.cz>
8367 PR tree-optimization/85799
8368 * passes.def: Add argument for pass_strip_predict_hints.
8369 * predict.c (class pass_strip_predict_hints): Add new argument
8371 (strip_predictor_early): New function.
8372 (pass_strip_predict_hints::execute): Call the function to
8374 (strip_predict_hints): New function.
8375 * predict.def: Fix comment.
8377 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
8379 * Makefile.in: Clarify which tm.texi to copy over to assert the
8380 right to grant a GFDL license for all.
8382 2018-08-09 Jeff Law <law@redhat.com>
8384 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
8387 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
8389 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
8392 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
8394 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
8395 reductions for variable-length vectors.
8397 2018-08-09 David Malcolm <dmalcolm@redhat.com>
8400 * common.opt (fdiagnostics-show-line-numbers): New option.
8401 * diagnostic-show-locus.c (class layout): Add fields
8402 "m_show_line_numbers_p" and "m_linenum_width";
8403 (num_digits): New function.
8404 (test_num_digits): New function.
8405 (layout::layout): Initialize new fields. Update m_x_offset
8406 logic to handle any left margin.
8407 (layout::print_source_line): Print line number when requested.
8408 (layout::start_annotation_line): New member function.
8409 (layout::print_annotation_line): Call it.
8410 (layout::print_leading_fixits): Likewise.
8411 (layout::print_trailing_fixits): Likewise. Update calls to
8412 move_to_column for new parameter.
8413 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
8414 it to potentially call start_annotation_line.
8415 (layout::show_ruler): Call start_annotation_line.
8416 (selftest::test_line_numbers_multiline_range): New selftest.
8417 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
8418 and selftest::test_line_numbers_multiline_range.
8419 * diagnostic.c (diagnostic_initialize): Initialize
8420 show_line_numbers_p.
8421 * diagnostic.h (struct diagnostic_context): Add field
8422 "show_line_numbers_p".
8423 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
8424 -fno-diagnostics-show-line-numbers.
8425 * dwarf2out.c (gen_producer_string): Add
8426 OPT_fdiagnostics_show_line_numbers to the ignored options.
8427 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
8428 one setting" options.
8429 (append_compiler_options): Likewise to the dropped options.
8430 (append_diag_options): Likewise to the passed-on options.
8431 * opts.c (common_handle_option): Handle the new option.
8432 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
8434 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
8436 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
8437 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
8438 third argument of type "const signed char" to descriptions of
8439 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
8440 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
8441 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
8442 __builtin_bcdsub_ov functions.
8444 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
8446 PR tree-optimization/86858
8447 * tree-vect-loop.c (vect_is_simple_reduction): Restore
8448 flow_bb_inside_loop_p calls.
8450 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
8452 PR tree-optimization/86871
8453 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
8454 instead of gimple_assign_lhs.
8456 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
8459 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
8460 register constraint to operand 0.
8461 (add<mode>3_carryinC): Likewise.
8462 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
8464 2018-08-09 Martin Liska <mliska@suse.cz>
8467 * common.opt: Remove extra line.
8469 2018-08-09 Martin Liska <mliska@suse.cz>
8471 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
8472 at the end of a line, make first letter capital and end up
8473 a sentence with a dot.
8474 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
8475 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
8476 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
8477 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
8478 (PARAM_MAX_ISL_OPERATIONS): Likewise.
8479 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
8480 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
8481 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
8482 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
8483 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
8484 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
8485 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
8486 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
8487 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
8488 (PARAM_TREE_REASSOC_WIDTH): Likewise.
8489 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
8490 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
8491 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
8493 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
8496 * config/s390/s390.c (s390_option_override_internal): Reduce the
8497 stack-clash-protection-probe-interval param if it would be too big
8500 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
8503 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
8504 * config/m68k/m68k.c (handle_move_double): Don't call
8505 m68k_final_prescan_insn.
8506 (m68k_adjust_decorated_operand): Renamed from
8507 m68k_final_prescan_insn, remove first and third operand and
8509 (print_operand): Call it.
8510 (print_operand_address): Call it.
8512 2018-08-08 Nathan Sidwell <nathan@acm.org>
8514 * diagnostic.c (diagnostic_report_current_module): Use
8515 linemap_included_from & linemap_included_from_linemap.
8517 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
8519 * config/aarch64/aarch64-cores.def: Add phecda core.
8520 * config/aarch64/aarch64-tune.md: Regenerate.
8521 * doc/invoke.texi: Add phecda core.
8523 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
8526 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
8528 * config/s390/s390.md ("movti"): Add more alternatives for
8529 constant to GPR copies.
8531 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
8533 * config/s390/s390.c: Fix whitespace damage throughout the file.
8534 * config/s390/s390.h: Likewise.
8535 * config/s390/tpf.h: Likewise.
8537 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
8539 * config/s390/s390.c (s390_loadrelative_operand_p):
8540 Remove TARGET_CPU_ZARCH usages.
8541 (s390_rtx_costs): Likewise.
8542 (s390_legitimate_constant_p): Likewise.
8543 (s390_cannot_force_const_mem): Likewise.
8544 (legitimate_reload_constant_p): Likewise.
8545 (s390_preferred_reload_class): Likewise.
8546 (legitimize_pic_address): Likewise.
8547 (legitimize_tls_address): Likewise.
8548 (s390_split_branches): Removed.
8549 (s390_add_execute): Removed.
8550 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
8551 (s390_mainpool_start): Likewise.
8552 (s390_mainpool_finish): Likewise.
8553 (s390_mainpool_cancel): Removed.
8554 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
8555 (s390_chunkify_cancel): Likewise.
8556 (s390_return_addr_rtx): Likewise.
8557 (s390_register_info): Remove split_branches_pending_p uages.
8558 (s390_optimize_register_info): Likewise.
8559 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
8560 split_branches_pending_p usages.
8561 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
8562 (s390_load_got): Likewise.
8563 (s390_expand_split_stack_prologue): Likewise.
8564 (output_asm_nops): Likewise.
8565 (s390_function_profiler): Likewise.
8566 (s390_emit_call): Likewise.
8567 (s390_conditional_register_usage): Likewise.
8568 (s390_optimize_prologue): Likewise.
8569 (s390_reorg): Remove TARGET_CPU_ZARCH and
8570 split_branches_pending_p usages.
8571 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
8573 (s390_output_indirect_thunk_function): Likewise.
8574 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
8575 (TARGET_CPU_ZARCH_P): Removed.
8576 (struct machine_function): Remove split_branches_pending_p.
8577 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
8579 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
8581 * common/config/s390/s390-common.c (processor_flags_table):
8583 * config.gcc: Remove with_arch/with_tune support.
8584 * config/s390/2064.md: Remove cpu attribute comparisons.
8585 * config/s390/driver-native.c (s390_host_detect_local_cpu):
8587 * config/s390/linux.h (ASM_SPEC):
8588 Remove -march support.
8589 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
8590 Use a table to get an arch level.
8591 * config/s390/s390-opts.h (enum processor_type):
8593 * config/s390/s390.c
8594 (processor_table): Remove entries, add arch_level values.
8595 (s390_issue_rate): Remove cases.
8596 (s390_option_override): Adjust
8597 s390_option_override_internal() call.
8598 (s390_option_override_internal): Remove deprecation warning.
8599 (s390_valid_target_attribute_tree): Adjust
8600 s390_option_override_internal() call.
8601 * config/s390/s390.h (struct s390_processor):
8602 Share with s390-c.c, add arch_level field.
8603 * config/s390/s390.md:
8604 Remove occurrences in cpu attribute.
8605 * config/s390/s390.opt: Remove -march/-mtune support.
8606 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
8607 * doc/invoke.texi: Remove deprecation warning.
8609 2018-08-08 Luis Machado <luis.machado@linaro.org>
8611 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
8613 (qdf24xx_tunings): Set vector cost structure to
8614 qdf24xx_vector_cost.
8616 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
8617 <register_sextend>: Set to 3.
8619 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
8622 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
8623 * config/aarch64/aarch64-simd.md
8624 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
8625 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
8626 (aarch64_frecpx<mode>): ...this new pattern.
8627 * config/aarch64/aarch64-simd-builtins.def: Remove comment
8628 about aarch64_frecp<FRECP:frecp_suffix><mode>.
8630 2018-08-07 Martin Liska <mliska@suse.cz>
8633 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
8634 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
8635 * predict.def (PRED_MALLOC_NONNULL): New predictor.
8636 * doc/extend.texi: Document that malloc attribute adds
8639 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
8642 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8643 Define to speculation_safe_value_not_needed.
8645 2018-08-06 Jeff Law <law@redhat.com>
8647 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
8648 the vr_values instance to cprop_into_stmt.
8649 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
8650 (cprop_operand): Also query EVRP to determine if OP is a constant.
8652 2018-08-06 Nathan Sidwell <nathan@acm.org>
8654 * diagnostic.c (diagnostic_report_current_module): Reroll
8655 included-at loop. Translate text.
8657 2018-08-06 David Malcolm <dmalcolm@redhat.com>
8659 * function-tests.c (selftest::test_expansion_to_rtl): Call
8660 free_after_compilation.
8662 2018-08-06 Alan Hayward <alan.hayward@arm.com>
8664 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
8666 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
8668 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
8669 loops with memory block operations from getting unrolled.
8671 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
8674 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8675 Define to speculation_safe_value_not_needed.
8677 2018-08-06 Jeff Law <law@redhat.com>
8679 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
8682 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8685 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
8686 with all enabled __intN types.
8688 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
8690 2018-08-06 Alan Hayward <alan.hayward@arm.com>
8692 * alias.c (record_set): Check for clobber high.
8693 * cfgexpand.c (expand_gimple_stmt): Likewise.
8694 * combine-stack-adj.c (single_set_for_csa): Likewise.
8695 * combine.c (find_single_use_1): Likewise.
8696 (set_nonzero_bits_and_sign_copies): Likewise.
8697 (get_combine_src_dest): Likewise.
8698 (is_parallel_of_n_reg_sets): Likewise.
8699 (try_combine): Likewise.
8700 (record_dead_and_set_regs_1): Likewise.
8701 (reg_dead_at_p_1): Likewise.
8702 (reg_dead_at_p): Likewise.
8703 * dce.c (deletable_insn_p): Likewise.
8704 (mark_nonreg_stores_1): Likewise.
8705 (mark_nonreg_stores_2): Likewise.
8706 * df-scan.c (df_find_hard_reg_defs): Likewise.
8707 (df_uses_record): Likewise.
8708 (df_get_call_refs): Likewise.
8709 * dwarf2out.c (mem_loc_descriptor): Likewise.
8710 * haifa-sched.c (haifa_classify_rtx): Likewise.
8711 * ira-build.c (create_insn_allocnos): Likewise.
8712 * ira-costs.c (scan_one_insn): Likewise.
8713 * ira.c (equiv_init_movable_p): Likewise.
8714 (rtx_moveable_p): Likewise.
8715 (interesting_dest_for_shprep): Likewise.
8716 * jump.c (mark_jump_label_1): Likewise.
8717 * postreload-gcse.c (record_opr_changes): Likewise.
8718 * postreload.c (reload_cse_simplify): Likewise.
8719 (struct reg_use): Add source expr.
8720 (reload_combine): Check for clobber high.
8721 (reload_combine_note_use): Likewise.
8722 (reload_cse_move2add): Likewise.
8723 (move2add_note_store): Likewise.
8724 * print-rtl.c (print_pattern): Likewise.
8725 * recog.c (decode_asm_operands): Likewise.
8726 (store_data_bypass_p): Likewise.
8727 (if_test_bypass_p): Likewise.
8728 * regcprop.c (kill_clobbered_value): Likewise.
8729 (kill_set_value): Likewise.
8730 * reginfo.c (reg_scan_mark_refs): Likewise.
8731 * reload1.c (maybe_fix_stack_asms): Likewise.
8732 (eliminate_regs_1): Likewise.
8733 (elimination_effects): Likewise.
8734 (mark_not_eliminable): Likewise.
8735 (scan_paradoxical_subregs): Likewise.
8736 (forget_old_reloads_1): Likewise.
8737 * reorg.c (find_end_label): Likewise.
8738 (try_merge_delay_insns): Likewise.
8739 (redundant_insn): Likewise.
8740 (own_thread_p): Likewise.
8741 (fill_simple_delay_slots): Likewise.
8742 (fill_slots_from_thread): Likewise.
8743 (dbr_schedule): Likewise.
8744 * resource.c (update_live_status): Likewise.
8745 (mark_referenced_resources): Likewise.
8746 (mark_set_resources): Likewise.
8747 * rtl.c (copy_rtx): Likewise.
8748 * rtlanal.c (reg_referenced_p): Likewise.
8749 (single_set_2): Likewise.
8750 (noop_move_p): Likewise.
8751 (note_stores): Likewise.
8752 * sched-deps.c (sched_analyze_reg): Likewise.
8753 (sched_analyze_insn): Likewise.
8755 2018-08-06 Alan Hayward <alan.hayward@arm.com>
8757 * cse.c (invalidate_reg): New function extracted from...
8758 (invalidate): ...here.
8759 (canonicalize_insn): Check for clobber high.
8760 (invalidate_from_clobbers): invalidate clobber highs.
8761 (invalidate_from_sets_and_clobbers): Likewise.
8762 (count_reg_usage): Check for clobber high.
8763 (insn_live_p): Likewise.
8764 * cselib.c (cselib_expand_value_rtx_1):Likewise.
8765 (cselib_invalidate_regno): Check for clobber in setter.
8766 (cselib_invalidate_rtx): Pass through setter.
8767 (cselib_invalidate_rtx_note_stores):
8768 (cselib_process_insn): Check for clobber high.
8769 * cselib.h (cselib_invalidate_rtx): Add operand.
8771 2018-08-06 Alan Hayward <alan.hayward@arm.com>
8773 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
8774 (mark_not_eliminable): Likewise.
8775 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
8776 * lra-lives.c (process_bb_lives): Check for clobber high.
8777 * lra.c (new_insn_reg): Remember clobber highs.
8778 (collect_non_operand_hard_regs): Check for clobber high.
8779 (lra_set_insn_recog_data): Likewise.
8780 (add_regs_to_insn_regno_info): Likewise.
8781 (lra_update_insn_regno_info): Likewise.
8783 2018-08-06 Alan Hayward <alan.hayward@arm.com>
8785 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
8786 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
8788 2018-08-06 Alan Hayward <alan.hayward@arm.com>
8790 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
8791 (copy_insn_1): Likewise.
8792 (gen_hard_reg_clobber_high): New gen function.
8793 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
8794 * genemit.c (gen_exp): Likewise.
8795 (gen_emit_seq): Pass through info.
8796 (gen_insn): Check for CLOBBER_HIGH.
8797 (gen_expand): Pass through info.
8798 (gen_split): Likewise.
8799 (output_add_clobbers): Likewise.
8800 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
8801 (remove_clobbers): Likewise.
8802 * rtl.h (gen_hard_reg_clobber_high): New declaration.
8804 2018-08-06 Alan Hayward <alan.hayward@arm.com>
8806 * doc/rtl.texi (clobber_high): Add.
8807 (parallel): Add in clobber high
8808 * rtl.c (rtl_check_failed_code3): Add function.
8809 * rtl.def (CLOBBER_HIGH): Add expression.
8810 * rtl.h (RTL_CHECKC3): Add macro.
8811 (rtl_check_failed_code3): Add declaration.
8812 (XC3EXP): Add macro.
8814 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
8817 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
8818 cfun->machine->max_used_stack_alignment if needed.
8820 2018-08-04 Martin Sebor <msebor@redhat.com>
8822 PR tree-optimization/86571
8823 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
8826 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
8829 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8832 2018-08-03 Jeff Law <law@redhat.com>
8835 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8836 Define to speculation_safe_value_not_needed.
8838 2018-08-03 David Malcolm <dmalcolm@redhat.com>
8840 * doc/gcov.texi (-x): Remove duplicate "to".
8841 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
8842 (-Wif-not-aligned): Remove duplicate "is".
8843 (-flto): Remove duplicate "the".
8844 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
8845 duplicate "v5.00.b".
8846 (MSP430 Options): Remove duplicate "and" from the description
8847 of "-mgprel-sec=regexp".
8848 (x86 Options): Remove duplicate copies of "vmldLog102" and
8849 vmlsLog104 from description of "-mveclibabi=type".
8851 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
8853 * internal-fn.h (first_commutative_argument): Declare.
8854 * internal-fn.c (first_commutative_argument): New function.
8855 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
8856 restrictions for pattern statements. Use first_commutative_argument
8857 to look for commutative operands in calls to internal functions.
8859 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
8861 * Makefile.in (wide-int-range.o): New.
8862 * tree-vrp.c: Move all the wide_int_* functions to...
8863 * wide-int-range.cc: ...here.
8864 * tree-vrp.h: Move all the wide_int_* prototypes to...
8865 * wide-int-range.h: ...here.
8867 2018-08-03 Tom de Vries <tdevries@suse.de>
8869 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
8871 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
8872 * except.c (output_function_exception_table): Do early exit if
8873 targetm_common.except_unwind_info (&global_options) == UI_NONE.
8875 2018-08-03 Martin Liska <mliska@suse.cz>
8877 * predict.c (dump_prediction): Change to 2 digits
8880 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
8882 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
8884 (falkor_am_1_gtov_gtov): ... a new insn reservation.
8886 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
8888 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
8889 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
8890 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
8891 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
8892 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
8894 2018-08-02 David Malcolm <dmalcolm@redhat.com>
8896 * diagnostic-show-locus.c (diagnostic_show_locus): Use
8897 pp_take_prefix when saving the existing prefix.
8898 * diagnostic.c (diagnostic_append_note): Likewise.
8899 * langhooks.c (lhd_print_error_function): Likewise.
8900 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
8901 param's type. Free the existing prefix.
8902 (pp_take_prefix): New function.
8903 (pretty_printer::pretty_printer): Drop the prefix parameter.
8904 Rename the length parameter to match the comment.
8905 (pretty_printer::~pretty_printer): Free the prefix.
8906 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
8908 (struct pretty_printer): Drop the "const" from "prefix" field's
8909 type and clarify memory management.
8910 (pp_set_prefix): Drop the "const" from the 2nd param.
8911 (pp_take_prefix): New decl.
8913 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
8915 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
8916 for word_mode_ok here instead of passing as argument.
8917 (expand_block_compare): Change select_block_compare_mode() call.
8918 (expand_strncmp_gpr_sequence): New function.
8919 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
8921 2018-08-02 Jeff Law <law@redhat.com>
8924 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8925 Define to speculation_safe_value_not_needed.
8928 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8929 Define to speculation_safe_value_not_needed.
8931 2018-08-02 Tom de Vries <tdevries@suse.de>
8934 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
8935 function. Return UI_TARGET unconditionally.
8936 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
8937 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
8939 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
8941 * genemit.c (print_overload_test): Fix typo.
8943 2018-08-02 Richard Biener <rguenther@suse.de>
8945 PR tree-optimization/86816
8946 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
8947 which checks for value availability before querying it.
8948 (gvn_uses_equal): Use it.
8949 (same_succ_hash): Likewise.
8950 (gimple_equal_p): Likewise.
8952 2018-08-02 Nick Clifton <nickc@redhat.com>
8955 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8956 Define to speculation_safe_value_not_needed.
8959 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8960 Define to speculation_safe_value_not_needed.
8963 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8964 Define to speculation_safe_value_not_needed.
8967 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8968 Define to speculation_safe_value_not_needed.
8971 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8972 Define to speculation_safe_value_not_needed.
8975 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8976 Define to speculation_safe_value_not_needed.
8979 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8980 Define to speculation_safe_value_not_needed.
8983 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8984 Define to speculation_safe_value_not_needed.
8987 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
8988 speculation_safe_value_not_needed.
8991 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
8992 to speculation_safe_value_not_needed.
8994 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
8996 * doc/md.texi: Expand the documentation of instruction names
8997 to mention port-local uses. Document '@' in pattern names.
8998 * read-md.h (overloaded_instance, overloaded_name): New structs.
9000 (md_reader::handle_overloaded_name): New member function.
9001 (md_reader::get_overloads): Likewise.
9002 (md_reader::m_first_overload): New member variable.
9003 (md_reader::m_next_overload_ptr): Likewise.
9004 (md_reader::m_overloads_htab): Likewise.
9005 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
9006 m_next_overload_ptr and m_overloads_htab.
9007 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
9008 (get_mode_token, get_code_token, get_int_token): New functions.
9009 (map_attr_string): Add an optional argument that passes back
9010 the associated iterator.
9011 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
9012 (md_reader::handle_overloaded_name, add_overload_instance): New
9014 (apply_iterators): Handle '@' names. Report an error if '@'
9015 is used without iterators.
9016 (initialize_iterators): Initialize the new iterator_group fields.
9017 * genopinit.c (handle_overloaded_code_for)
9018 (handle_overloaded_gen): New functions.
9019 (main): Use them to print declarations of maybe_code_for_* and
9020 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
9021 * genemit.c (print_overload_arguments, print_overload_test)
9022 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
9023 (main): Use it to print definitions of maybe_code_for_* and
9024 maybe_gen_* functions.
9025 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
9026 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
9027 instead of explicit mode checks.
9028 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
9029 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
9030 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
9031 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
9032 (aarch64_expand_compare_and_swap): Likewise
9033 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
9034 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
9035 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
9036 (aarch64_constant_pool_reload_icode): Delete.
9037 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
9038 instead of aarch64_constant_pool_reload_icode. Use
9039 code_for_aarch64_reload_mov instead of explicit mode checks.
9040 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
9041 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
9042 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
9043 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
9044 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
9045 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
9046 (aarch64_atomic_load_op_code): Delete.
9047 (aarch64_emit_atomic_load_op): Likewise.
9048 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
9049 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
9050 instead of aarch64_emit_atomic_load_op.
9051 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
9052 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
9053 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
9054 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
9055 character before the pattern name.
9056 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
9057 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
9058 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
9059 (aarch64_frecps<mode>): Likewise.
9060 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
9061 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
9062 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
9063 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
9064 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
9066 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
9068 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
9069 Allow HFmode constants if TARGET_FP_F16INST.
9071 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
9074 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
9075 No longer check last store for clobber of address register.
9077 2018-08-02 Martin Liska <mliska@suse.cz>
9079 PR gcov-profile/86817
9080 * gcov.c (process_all_functions): New function.
9082 (process_file): Move functions processing to
9083 process_all_functions.
9085 2018-08-02 David Malcolm <dmalcolm@redhat.com>
9087 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
9088 "const" to the "gimple *" and "rtx_insn *" parameters.
9089 * dumpfile.h (dump_user_location_t::dump_user_location_t):
9091 (dump_location_t::dump_location_t): Likewise.
9093 2018-08-01 Martin Sebor <msebor@redhat.com>
9095 PR tree-optimization/86650
9096 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
9097 rather than a "gcall *". Directly pass the data of interest
9098 to percent_K_format, rather than building a temporary CALL_EXPR
9100 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
9101 (gimple_fold_builtin_strncat): Adjust.
9102 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
9103 gcall* argument with gimple*.
9104 * gimple-ssa-warn-restrict.c (check_call): Same.
9105 (wrestrict_dom_walker::before_dom_children): Same.
9106 (builtin_access::builtin_access): Same.
9107 (check_bounds_or_overlap): Same
9108 (maybe_diag_overlap): Same.
9109 (maybe_diag_offset_bounds): Same.
9110 * tree-diagnostic.c (default_tree_printer): Move usage of
9111 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
9113 * tree-pretty-print.c (percent_K_format): Add argument.
9114 * tree-pretty-print.h: Add argument.
9115 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
9116 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
9117 (maybe_diag_stxncpy_trunc): Same.
9118 (handle_builtin_stxncpy): Same.
9119 (handle_builtin_strcat): Same.
9121 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
9123 * match.pd: Optimise pointer range checks.
9125 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
9127 PR tree-optimization/86758
9128 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
9129 to remove pattern statements.
9131 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
9133 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
9134 result of dfs_enumerate_from when constructing stmt_vec_infos,
9135 instead of additionally calling get_loop_body.
9137 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
9139 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
9141 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
9142 When creating an iv, assert that the step is not known to be zero.
9143 (vect_setup_realignment): Update call accordingly.
9144 * tree-vect-stmts.c (vectorizable_store): Likewise.
9145 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
9147 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
9149 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
9150 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
9151 (vectorizable_reduction): Likewise.
9152 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9153 (vect_detect_hybrid_slp_stmts): Likewise.
9154 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9156 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
9158 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
9159 (wide_int_set_zero_nonzero_bits): ...this.
9160 (zero_nonzero_bits_from_vr): Rename to...
9161 (vrp_set_zero_nonzero_bits): ...this.
9162 (extract_range_from_multiplicative_op_1): Abstract wide int
9164 (wide_int_range_multiplicative_op): ...here.
9165 (extract_range_from_binary_expr_1): Extract wide int binary
9166 operations into their own functions.
9167 (wide_int_range_lshift): New.
9168 (wide_int_range_can_optimize_bit_op): New.
9169 (wide_int_range_shift_undefined_p): New.
9170 (wide_int_range_bit_xor): New.
9171 (wide_int_range_bit_ior): New.
9172 (wide_int_range_bit_and): New.
9173 (wide_int_range_trunc_mod): New.
9174 (extract_range_into_wide_ints): New.
9175 (vrp_shift_undefined_p): New.
9176 (extract_range_from_multiplicative_op): New.
9177 (vrp_can_optimize_bit_op): New.
9178 * tree-vrp.h (value_range::dump): New.
9179 (wide_int_range_multiplicative_op): New.
9180 (wide_int_range_lshift):New.
9181 (wide_int_range_shift_undefined_p): New.
9182 (wide_int_range_bit_xor): New.
9183 (wide_int_range_bit_ior): New.
9184 (wide_int_range_bit_and): New.
9185 (wide_int_range_trunc_mod): New.
9186 (zero_nonzero_bits_from_bounds): Rename to...
9187 (wide_int_set_zero_nonzero_bits): ...this.
9188 (zero_nonzero_bits_from_vr): Rename to...
9189 (vrp_set_zero_nonzero_bits): ...this.
9190 (range_easy_mask_min_max): Rename to...
9191 (wide_int_range_can_optimize_bit_op): this.
9192 * vr-values.c (simplify_bit_ops_using_ranges): Rename
9193 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
9195 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
9197 * tree-vectorizer.h (vect_orig_stmt): New function.
9198 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
9199 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
9200 (vect_create_epilog_for_reduction): Likewise.
9201 (vectorizable_live_operation): Likewise.
9202 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
9203 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
9204 * tree-vect-stmts.c (vectorizable_call): Likewise.
9205 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
9207 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
9209 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
9211 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
9212 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
9213 (vect_transform_loop): Likewise.
9214 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
9216 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
9218 * tree-vectorizer.h (vect_schedule_slp): Return void.
9219 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
9220 (vect_schedule_slp): Likewise.
9222 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
9224 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
9226 (vect_transform_loop): Update calls accordingly. Schedule SLP
9227 instances before the main loop, if any exist.
9229 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
9231 PR tree-optimization/86749
9232 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
9233 If the lhs is used in a COND_EXPR, check that it is being used
9234 as the "then" or "else" value.
9236 2018-08-01 Tom de Vries <tdevries@suse.de>
9239 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
9240 speculation_safe_value_not_needed.
9242 2018-08-01 Richard Biener <rguenther@suse.de>
9244 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
9247 2018-08-01 Martin Liska <mliska@suse.cz>
9249 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
9250 format how successful transformation is dumped.
9251 (gimple_mod_pow2_value_transform): Likewise.
9252 (gimple_mod_subtract_transform): Likewise.
9253 (gimple_stringops_transform): Likewise.
9255 2018-08-01 Martin Liska <mliska@suse.cz>
9258 * value-prof.c (interesting_stringop_to_profile_p):
9259 Simplify the code and add BUILT_IN_MEMMOVE.
9260 (gimple_stringops_transform): Likewise.
9262 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
9264 * config/aarch64/aarch64-simd.md
9265 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
9266 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
9267 use GPI iterator instead of SI mode.
9269 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
9271 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
9272 rs6000_speculation_barrier.
9273 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
9274 new barrier pattern name.
9276 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
9278 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
9279 (speculation_barrier): New insn.
9281 2018-08-01 Richard Biener <rguenther@suse.de>
9284 * graphite.h: Include isl/id.h and isl/space.h to allow build
9287 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
9290 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
9291 mode for COFF targets.
9292 * defaults.h (TARGET_COFF): Define.
9293 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
9294 TARGET_COFF): Define.
9295 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
9296 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
9297 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
9299 2018-07-31 Alexandre Oliva <oliva@adacore.com>
9301 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
9302 * gimple-streamer-out.c (output_bb): Save it.
9303 * lto-streamer-in.c (input_struct_function_base): Restore
9304 instance discriminator if available. Create map on demand.
9305 * lto-streamer-out.c (output_struct_function_base): Save it if
9307 * final.c (decl_to_instance_map): Document LTO strategy.
9309 2018-07-31 Alexandre Oliva <oliva@adacore.com>
9310 Olivier Hainque <hainque@adacore.com>
9312 * debug.h (decl_to_instance_map_t): New type.
9313 (decl_to_instance_map): Declare.
9314 (maybe_create_decl_to_instance_map): New inline function.
9315 * final.c (bb_discriminator, last_bb_discriminator): New statics,
9316 to track basic block discriminators.
9317 (final_start_function_1): Initialize them.
9318 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
9320 (decl_to_instance_map): New variable.
9321 (map_decl_to_instance, maybe_set_discriminator): New functions.
9322 (notice_source_line): Set discriminator.
9324 2018-07-31 Ian Lance Taylor <iant@golang.org>
9326 * targhooks.c (default_have_speculation_safe_value): Add
9329 2018-07-31 David Malcolm <dmalcolm@redhat.com>
9331 * dump-context.h: Include "pretty-print.h".
9332 (dump_context::refresh_dumps_are_enabled): New decl.
9333 (dump_context::emit_item): New decl.
9334 (class dump_context): Add fields "m_test_pp" and
9336 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
9337 (temp_dump_context::get_dumped_text): New decl.
9338 (class temp_dump_context): Add field "m_pp".
9339 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
9340 (dump_context::refresh_dumps_are_enabled): ...and add a test for
9342 (set_dump_file): Update for above change.
9343 (set_alt_dump_file): Likewise.
9344 (dump_loc): New overload, taking a pretty_printer *.
9345 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
9346 to any test pretty-printer.
9347 (make_item_for_dump_gimple_stmt): New function, adapted from
9348 optinfo::add_gimple_stmt.
9349 (dump_context::dump_gimple_stmt): Call it, and use the result,
9350 eliminating the direct usage of dump_file and alt_dump_file in
9351 favor of indirectly using them via emit_item.
9352 (make_item_for_dump_gimple_expr): New function, adapted from
9353 optinfo::add_gimple_expr.
9354 (dump_context::dump_gimple_expr): Call it, and use the result,
9355 eliminating the direct usage of dump_file and alt_dump_file in
9356 favor of indirectly using them via emit_item.
9357 (make_item_for_dump_generic_expr): New function, adapted from
9359 (dump_context::dump_generic_expr): Call it, and use the result,
9360 eliminating the direct usage of dump_file and alt_dump_file in
9361 favor of indirectly using them via emit_item.
9362 (make_item_for_dump_printf_va): New function, adapted from
9363 optinfo::add_printf_va.
9364 (make_item_for_dump_printf): New function.
9365 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
9366 and use the result, eliminating the direct usage of dump_file and
9367 alt_dump_file in favor of indirectly using them via emit_item.
9368 (make_item_for_dump_dec): New function.
9369 (dump_context::dump_dec): Call it, and use the result,
9370 eliminating the direct usage of dump_file and alt_dump_file in
9371 favor of indirectly using them via emit_item.
9372 (make_item_for_dump_symtab_node): New function, adapted from
9373 optinfo::add_symtab_node.
9374 (dump_context::dump_symtab_node): Call it, and use the result,
9375 eliminating the direct usage of dump_file and alt_dump_file in
9376 favor of indirectly using them via emit_item.
9377 (dump_context::begin_scope): Reimplement, avoiding direct usage
9378 of dump_file and alt_dump_file in favor of indirectly using them
9380 (dump_context::emit_item): New member function.
9381 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
9382 Set up test pretty-printer on the underlying context. Call
9383 refresh_dumps_are_enabled.
9384 (temp_dump_context::~temp_dump_context): Call
9385 refresh_dumps_are_enabled.
9386 (temp_dump_context::get_dumped_text): New member function.
9387 (selftest::verify_dumped_text): New function.
9388 (ASSERT_DUMPED_TEXT_EQ): New macro.
9389 (selftest::test_capture_of_dump_calls): Run all tests twice, with
9390 and then without optinfo enabled. Add uses of
9391 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
9392 * dumpfile.h: Update comment for the dump_* API.
9393 * optinfo-emit-json.cc
9394 (selftest::test_building_json_from_dump_calls): Update for new
9395 param for temp_dump_context ctor.
9396 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
9397 and "m_owned" field.
9398 (optinfo_item::~optinfo_item): Likewise.
9399 (optinfo::add_item): New member function.
9400 (optinfo::emit): Update comment.
9401 (optinfo::add_string): Delete.
9402 (optinfo::add_printf): Delete.
9403 (optinfo::add_printf_va): Delete.
9404 (optinfo::add_gimple_stmt): Delete.
9405 (optinfo::add_gimple_expr): Delete.
9406 (optinfo::add_tree): Delete.
9407 (optinfo::add_symtab_node): Delete.
9408 (optinfo::add_dec): Delete.
9409 * optinfo.h (class dump_context): New forward decl.
9410 (optinfo::add_item): New decl.
9411 (optinfo::add_string): Delete.
9412 (optinfo::add_printf): Delete.
9413 (optinfo::add_printf_va): Delete.
9414 (optinfo::add_gimple_stmt): Delete.
9415 (optinfo::add_gimple_expr): Delete.
9416 (optinfo::add_tree): Delete.
9417 (optinfo::add_symtab_node): Delete.
9418 (optinfo::add_dec): Delete.
9419 (optinfo::add_poly_int): Delete.
9420 (optinfo_item::optinfo_item): Remove "owned" param.
9421 (class optinfo_item): Remove field "m_owned".
9423 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9426 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
9427 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
9428 requested variable alignment.
9429 (expand_one_ssa_partition): Likewise.
9430 (expand_one_var): Likewise.
9432 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
9434 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
9435 to speculation_safe_value_not_needed.
9437 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
9439 * targhooks.h (speculation_safe_value_not_needed): New prototype.
9440 * targhooks.c (speculation_safe_value_not_needed): New function.
9441 * target.def (have_speculation_safe_value): Update documentation.
9442 * doc/tm.texi: Regenerated.
9444 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
9446 * config/aarch64/iterators.md (ALLI_TI): New iterator.
9447 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
9449 (despeculate_copy<ALLI:mode>_insn): New insn.
9450 (despeculate_copyti_insn): New insn.
9451 (despeculate_simple<ALLI:mode>): New insn
9452 (despeculate_simpleti): New insn.
9453 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
9455 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
9456 aarch64_speculation_safe_value.
9457 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
9459 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
9461 * config/aarch64/aarch64-speculation.cc: New file.
9462 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
9463 before pass_reorder_blocks.
9464 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
9466 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
9467 X14 and X15 when tracking speculation.
9468 * config/aarch64/aarch64.md (register name constants): Add
9469 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
9470 (unspec): Add UNSPEC_SPECULATION_TRACKER.
9471 (speculation_barrier): New insn attribute.
9472 (cmp<mode>): Allow SP in comparisons.
9473 (speculation_tracker): New insn.
9474 (speculation_barrier): Add speculation_barrier attribute.
9475 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
9476 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
9477 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
9479 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
9481 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
9482 aarch64_track_speculation is true.
9483 (tb<optab><mode>1): Likewise.
9484 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
9485 generate CB[N]Z when tracking speculation.
9486 (aarch64_split_compare_and_swap): Likewise.
9487 (aarch64_split_atomic_op): Likewise.
9489 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
9491 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
9493 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
9495 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
9496 (speculation_barrier): New insn.
9498 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
9500 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
9501 * config/arm/arm.md (speculation_barrier): New expand.
9502 (speculation_barrier_insn): New pattern.
9504 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
9506 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
9507 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
9508 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
9509 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
9511 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
9512 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
9513 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
9514 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
9515 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
9516 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
9517 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
9518 * builtins.c (expand_speculation_safe_value): New function.
9519 (expand_builtin): Call it.
9520 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
9521 * doc/extend.texi: Document __builtin_speculation_safe_value.
9522 * doc/md.texi: Document "speculation_barrier" pattern.
9523 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
9524 TARGET_HAVE_SPECULATION_SAFE_VALUE.
9525 * doc/tm.texi: Regenerated.
9526 * target.def (have_speculation_safe_value, speculation_safe_value): New
9528 * targhooks.c (default_have_speculation_safe_value): New function.
9529 (default_speculation_safe_value): New function.
9530 * targhooks.h (default_have_speculation_safe_value): Add prototype.
9531 (default_speculation_safe_value): Add prototype.
9533 2018-07-31 David Malcolm <dmalcolm@redhat.com>
9535 * dump-context.h (dump_context::dump_loc): New decl.
9536 * dumpfile.c (dump_context::dump_loc): New member function.
9537 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
9538 and dump_gimple_stmt.
9539 (dump_context::dump_gimple_expr_loc): Likewise, using
9541 (dump_context::dump_generic_expr_loc): Likewise, using
9543 (dump_context::dump_printf_loc_va): Likewise, using
9545 (dump_context::begin_scope): Explicitly using the global function
9546 "dump_loc", rather than the member function.
9548 2018-07-31 Martin Sebor <msebor@redhat.com>
9550 PR tree-optimization/86741
9551 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
9553 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
9555 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
9556 depend on whether prefetch instructions will be emitted or not.
9557 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
9558 will be emitted or not.
9559 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
9560 (TARGET_SETMEM_PFD): New macros.
9562 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9564 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
9565 (NULL_STMT_VEC_INFO): Delete.
9566 (stmt_vec_info::operator*): Likewise.
9567 (stmt_vec_info::operator gimple *): Likewise.
9568 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
9569 of NULL_STMT_VEC_INFO.
9570 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9571 (vect_reassociating_reduction_p): Likewise.
9572 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
9573 (vectorizable_store): Likewise.
9574 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
9575 (vec_info::free_stmt_vec_infos): Likewise.
9577 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9579 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
9580 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
9582 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9584 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
9585 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
9586 (vec_info::free_stmt_vec_info): New private member functions.
9587 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
9588 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
9589 * tree-parloops.c (gather_scalar_reductions): Remove calls to
9590 set_stmt_vec_info_vec and free_stmt_vec_infos.
9591 * tree-vect-loop.c (_loop_vec_info): Remove call to
9592 set_stmt_vec_info_vec.
9593 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
9594 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
9595 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
9596 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
9597 (vec_info::free_stmt_vec_info): ...these new functions. Remove
9598 assignments in {vec_info::,}new_stmt_vec_info that are redundant
9599 with the clearing in the xcalloc.
9600 (stmt_vec_info_vec): Delete.
9601 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
9602 (vectorize_loops): Likewise.
9603 (vec_info::~vec_info): Remove argument from call to
9604 free_stmt_vec_infos.
9605 (vec_info::add_stmt): Remove vinfo argument from call to
9608 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9610 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
9611 rather than a gimple stmt.
9612 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
9613 information for pattern statements when passed the original
9614 statement; instead wait to be passed the pattern statement itself.
9615 Don't call set_vinfo_for_stmt here.
9616 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
9617 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
9618 stmt_vec_infos here.
9619 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
9620 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
9621 stmt_vec_infos entry.
9623 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9625 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
9626 * tree-vectorizer.c (vec_info::replace_stmt): New function.
9627 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
9628 * tree-vect-stmts.c (vectorizable_call): Likewise.
9629 (vectorizable_simd_clone_call): Likewise.
9631 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9633 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
9634 * tree-vectorizer.c (vec_info::remove_stmt): New function.
9635 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
9636 * tree-vect-loop.c (vect_transform_loop): Likewise.
9637 * tree-vect-slp.c (vect_schedule_slp): Likewise.
9638 * tree-vect-stmts.c (vect_remove_stores): Likewise.
9640 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9642 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
9643 (vect_dr_stmt): Delete.
9644 * tree-vectorizer.c (vec_info::lookup_dr): New function.
9645 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
9647 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9648 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
9649 (vect_verify_datarefs_alignment, vect_peeling_supportable)
9650 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
9651 (vect_analyze_data_refs): Likewise.
9652 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
9654 (vect_find_same_alignment_drs): Likewise.
9655 (vect_slp_analyze_node_dependences): Update calls accordingly.
9656 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
9657 instead of DR_VECT_AUX.
9658 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
9659 of a vector data references. Use vec_info::lookup_dr instead of
9661 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
9662 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
9663 instead of DR_VECT_AUX.
9665 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9667 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
9669 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
9671 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
9672 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
9673 (vect_gen_prolog_loop_niters): Likewise.
9675 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9677 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
9678 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
9679 (vect_known_alignment_in_bytes, vect_dr_behavior)
9680 (vect_get_scalar_dr_size): Take references as dr_vec_infos
9681 instead of data_references. Update calls to other routines for
9682 which the same change has been made.
9683 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
9684 dr_vec_infos instead of stmt_vec_infos.
9685 (vect_analyze_data_ref_dependence): Update call accordingly.
9686 (vect_slp_analyze_data_ref_dependence)
9687 (vect_record_base_alignments): Use DR_VECT_AUX.
9688 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
9689 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
9690 (vector_alignment_reachable_p, vect_get_data_access_cost)
9691 (vect_peeling_supportable, vect_analyze_group_access_1)
9692 (vect_analyze_group_access, vect_analyze_data_ref_access)
9693 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
9694 (vect_compile_time_alias, vect_small_gap_p)
9695 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
9696 (vect_supportable_dr_alignment): Take references as dr_vec_infos
9697 instead of data_references. Update calls to other routines for
9698 which the same change has been made.
9699 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
9700 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
9701 (vect_slp_analyze_and_verify_node_alignment)
9702 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
9703 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
9704 (vect_setup_realignment): Use dr_vec_infos. Update calls after
9706 (_vect_peel_info::dr): Replace with...
9707 (_vect_peel_info::dr_info): ...this new field.
9708 (vect_peeling_hash_get_most_frequent)
9709 (vect_peeling_hash_choose_best_peeling): Update accordingly.
9710 (vect_peeling_hash_get_lowest_cost):
9711 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
9712 routines for which the same change has been made.
9713 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
9715 * tree-vect-loop-manip.c (get_misalign_in_elems)
9716 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
9718 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
9719 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
9720 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
9721 (get_group_load_store_type, get_negative_load_store_type)
9722 (vect_get_data_ptr_increment, vectorizable_store)
9723 (vectorizable_load): Likewise.
9724 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
9725 Update calls to other routines for which the same change has been made.
9727 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9729 * tree-vectorizer.h (vec_info::move_dr): New member function.
9730 (dataref_aux): Rename to...
9731 (dr_vec_info): ...this and add "dr" and "stmt" fields.
9732 (_stmt_vec_info::dr_aux): Update accordingly.
9733 (_stmt_vec_info::data_ref_info): Delete.
9734 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
9735 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
9736 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
9737 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
9739 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
9740 (STMT_VINFO_DR_INFO): New macro.
9741 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
9742 (set_dr_misalignment): Update after rename of dataref_aux.
9743 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
9744 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
9745 initialization of STMT_VINFO_DATA_REF.
9746 * tree-vectorizer.c (vec_info::move_dr): New function.
9747 * tree-vect-patterns.c (vect_recog_bool_pattern)
9748 (vect_recog_mask_conversion_pattern)
9749 (vect_recog_gather_scatter_pattern): Use it.
9750 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
9751 the "dr" and "stmt" fields of dr_vec_info instead of
9752 STMT_VINFO_DATA_REF.
9754 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9756 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
9757 (is_pattern_stmt_p): Use it.
9758 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
9759 on pattern statements.
9761 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9763 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
9764 original stmt as a stmt_vec_info rather than a gimple stmt.
9765 (vect_pattern_recog_1): Take the statement directly as a
9766 stmt_vec_info, rather than via a gimple_stmt_iterator.
9767 Update call to vect_mark_pattern_stmts.
9768 (vect_pattern_recog): Update calls accordingly.
9770 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9772 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
9773 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
9774 a vect_def_type for the first argument.
9775 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
9776 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
9777 operand if it isn't defined by a vectorized statement.
9778 (vect_build_gather_load_calls): Remove the mask_dt argument and
9779 update calls to vect_get_vec_def_for_stmt_copy.
9780 (vectorizable_bswap): Likewise the dt argument.
9781 (vectorizable_call): Update calls to vectorizable_bswap and
9782 vect_get_vec_def_for_stmt_copy.
9783 (vectorizable_simd_clone_call, vectorizable_assignment)
9784 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
9785 (vectorizable_comparison): Update calls to
9786 vect_get_vec_def_for_stmt_copy.
9787 (vectorizable_store): Likewise. Remove now-unnecessary calls to
9789 (vect_get_loop_based_defs): Remove dt argument and update call
9790 to vect_get_vec_def_for_stmt_copy.
9791 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
9792 and vect_get_vec_def_for_stmt_copy.
9793 (vectorizable_load): Update calls to vect_build_gather_load_calls
9794 and vect_get_vec_def_for_stmt_copy.
9795 * tree-vect-loop.c (vect_create_epilog_for_reduction)
9796 (vectorizable_reduction, vectorizable_live_operation): Update calls
9797 to vect_get_vec_def_for_stmt_copy.
9799 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9801 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
9802 and gimple stmt arguments with a stmt_vec_info.
9803 (vect_record_base_alignments): Update calls accordingly.
9804 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
9805 and gimple stmt arguments with a stmt_vec_info.
9806 (vect_build_slp_tree_1): Remove vinfo argument and update call
9807 to vect_record_max_nunits.
9808 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
9809 and vect_record_max_nunits.
9811 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9813 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
9814 file and take a stmt_vec_info instead of a gimple stmt.
9815 (supportable_widening_operation, vect_finish_replace_stmt)
9816 (vect_finish_stmt_generation, vect_get_store_rhs)
9817 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
9818 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
9819 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
9820 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
9821 (vect_create_data_ref_ptr, bump_vector_ptr)
9822 (vect_permute_store_chain, vect_setup_realignment)
9823 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
9824 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
9825 (vectorizable_reduction, vectorizable_induction)
9826 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
9827 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
9828 than gimple stmts as arguments.
9829 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
9830 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
9831 (can_group_stmts_p, vect_check_gather_scatter)
9832 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
9833 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
9834 (vect_permute_load_chain, vect_shift_permute_load_chain)
9835 (vect_transform_grouped_load)
9836 (vect_record_grouped_load_vectors): Likewise.
9837 * tree-vect-loop.c (vect_fixup_reduc_chain)
9838 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
9839 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
9840 (vectorizable_reduction, vectorizable_induction)
9841 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
9842 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
9843 (vect_get_load_store_mask): Likewise.
9844 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
9845 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
9846 * tree-vect-stmts.c (vect_mark_relevant)
9847 (is_simple_and_all_uses_invariant)
9848 (exist_non_indexing_operands_for_use_p, process_use)
9849 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
9850 (vect_get_vec_def_for_operand, vect_get_vec_defs)
9851 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
9852 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
9853 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
9854 (get_negative_load_store_type, get_load_store_type)
9855 (vect_check_load_store_mask, vect_check_store_rhs)
9856 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
9857 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
9858 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
9859 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9860 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
9861 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
9862 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
9863 (supportable_widening_operation): Likewise.
9865 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9867 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
9868 a stmt_vec_info instead of a gcall.
9869 (vect_check_gather_scatter): Update call accordingly.
9870 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
9872 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
9873 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
9874 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
9875 instead of a gimple stmt.
9876 (vect_transform_loop): Update calls accordingly.
9877 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
9878 stmt_vec_infos instead of gimple stmts.
9879 (vect_analyze_slp_instance): Update use accordingly.
9880 * tree-vect-stmts.c (read_vector_array, write_vector_array)
9881 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
9882 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
9883 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
9884 (vect_gen_widened_results_half, vect_get_loop_based_defs)
9885 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
9886 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
9887 down to subroutines.
9889 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9891 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
9892 of the worklist from a vector of gimple stmts to a vector of
9894 * tree-vect-stmts.c (vect_mark_relevant, process_use)
9895 (vect_mark_stmts_to_be_vectorized): Likewise
9897 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9899 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
9900 statement before passing it to vect_analyze_stmt.
9901 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
9902 the chain of phi vector definitions. Track the exit phi via its
9904 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
9905 STMT_VINFO_REDUC_DEF.
9906 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
9907 stmt_vec_infos to handle the statement chains.
9908 (vect_get_slp_defs): Record the first statement in the node
9909 using a stmt_vec_info.
9910 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
9911 statements here and pass their stmt_vec_info down to subroutines.
9912 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
9913 down to vect_finish_stmt_generation.
9914 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
9915 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
9916 stmt_vec_infos to subroutines.
9917 (vect_remove_stores): Use stmt_vec_infos to handle the statement
9920 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9922 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
9923 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
9924 (vect_permute_store_chain, vect_setup_realignment)
9925 (vect_permute_load_chain, vect_shift_permute_load_chain)
9926 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
9927 stmts internally, and when passing values to other vectorizer routines.
9928 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
9929 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
9930 (vect_analyze_loop_operations, get_initial_def_for_reduction)
9931 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
9932 (vectorizable_reduction, vectorizable_induction)
9933 (vectorizable_live_operation, vect_transform_loop_stmt)
9934 (vect_transform_loop): Likewise.
9935 * tree-vect-patterns.c (vect_reassociating_reduction_p)
9936 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
9937 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
9938 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9939 (vect_slp_analyze_node_operations_1): Likewise.
9940 * tree-vect-stmts.c (vect_mark_relevant, process_use)
9941 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
9942 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
9943 (vect_finish_stmt_generation_1, get_group_load_store_type)
9944 (get_load_store_type, vect_build_gather_load_calls)
9945 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
9946 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
9947 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9948 (vectorizable_store, vectorizable_load, vectorizable_condition)
9949 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
9950 (supportable_widening_operation): Likewise.
9951 (vect_get_vector_types_for_stmt): Likewise.
9952 * tree-vectorizer.h (vect_dr_behavior): Likewise.
9954 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9956 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9957 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
9958 (vect_permute_store_chain, vect_permute_load_chain)
9959 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
9960 repeated stmt_vec_info lookups.
9961 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
9962 (vect_update_ivs_after_vectorizer): Likewise.
9963 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
9964 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9965 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
9966 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9967 (vect_bb_slp_scalar_cost): Likewise.
9968 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
9970 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9972 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
9973 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
9974 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
9975 to get gassigns and gcalls, rather than operating on generc gimple
9977 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
9978 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
9979 (vectorizable_load, vect_analyze_stmt): Likewise.
9980 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
9982 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9984 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
9985 return stmt_vec_infos rather than gimple stmts. Do not accept
9987 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
9989 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
9990 Update use of get_later_stmt.
9991 (vect_get_constant_vectors): Update call accordingly.
9992 (vect_schedule_slp_instance): Likewise
9993 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
9994 (vect_slp_analyze_instance_dependence): Likewise.
9995 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
9997 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
9999 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
10000 (stmt_info_for_cost::stmt_info): ...this new field.
10001 (add_stmt_costs): Update accordingly.
10002 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
10003 (vect_get_known_peeling_cost): Likewise.
10004 (vect_estimate_min_profitable_iters): Likewise.
10005 * tree-vect-stmts.c (record_stmt_cost): Likewise.
10007 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10009 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
10010 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10011 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10013 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
10015 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10017 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
10018 a gimple stmt to a stmt_vec_info.
10019 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
10021 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10023 * tree-vectorizer.h (vec_info::grouped_stores): Change from
10024 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10025 (_loop_vec_info::reduction_chains): Likewise.
10026 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
10028 * tree-vect-slp.c (vect_analyze_slp): Likewise.
10030 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10032 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
10033 a gimple stmt to a stmt_vec_info.
10034 (_stmt_vec_info::next_element): Likewise.
10035 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
10036 (vect_slp_analyze_and_verify_node_alignment)
10037 (vect_analyze_group_access_1, vect_analyze_group_access)
10038 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
10039 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
10040 (vect_supportable_dr_alignment): Update accordingly.
10041 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
10042 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
10043 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
10044 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
10045 * tree-vect-slp.c (vect_build_slp_tree_1)
10046 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
10047 (vect_split_slp_store_group, vect_analyze_slp_instance)
10048 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
10049 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
10050 (get_group_load_store_type, get_load_store_type)
10051 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
10052 (vect_transform_stmt, vect_remove_stores): Likewise.
10054 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10056 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
10057 than a gimple stmt.
10058 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10059 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
10060 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
10061 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
10062 (vector_alignment_reachable_p, vect_get_data_access_cost)
10063 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
10064 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
10065 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
10066 (vect_analyze_group_access_1, vect_analyze_group_access)
10067 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
10068 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
10069 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
10070 result of vect_dr_stmt and use the stmt_vec_info instead of
10071 the associated gimple stmt.
10072 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10073 (vect_gen_prolog_loop_niters): Likewise.
10074 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10076 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10078 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
10079 to a vec<stmt_vec_info>.
10080 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
10081 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
10082 vec<stmt_vec_info>.
10083 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
10084 to a vec<stmt_vec_info>.
10085 (bst_traits::value_type, bst_traits::value_type): Likewise.
10086 (bst_traits::hash): Update accordingly.
10087 (vect_get_and_check_slp_defs): Change the stmts parameter from
10088 a vec<gimple *> to a vec<stmt_vec_info>.
10089 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
10090 (vect_build_slp_tree): Likewise.
10091 (vect_build_slp_tree_2): Likewise. Update uses of
10092 SLP_TREE_SCALAR_STMTS.
10093 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
10094 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
10095 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
10096 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
10097 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
10098 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
10099 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
10100 (vect_get_constant_vectors, vect_get_slp_defs)
10101 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
10102 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
10103 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
10104 instead of gimple stmts.
10105 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
10106 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
10107 (vect_slp_analyze_instance_dependence): Update uses of
10108 SLP_TREE_SCALAR_STMTS.
10109 (vect_slp_analyze_and_verify_node_alignment): Likewise.
10110 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
10111 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
10112 (get_initial_defs_for_reduction): Likewise.
10113 (vect_create_epilog_for_reduction): Likewise.
10114 (vectorize_fold_left_reduction): Likewise.
10115 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
10116 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
10117 (can_vectorize_live_stmts): Likewise.
10119 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10121 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
10122 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10123 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
10125 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
10127 (gather_scalar_reductions): Update after above interface changes.
10128 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
10129 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
10131 (vect_force_simple_reduction): Likewise.
10132 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
10133 LOOP_VINFO_REDUCTIONS.
10134 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10136 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10138 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
10139 a gimple stmt to a stmt_vec_info.
10140 * tree-vect-loop.c (vect_active_double_reduction_p)
10141 (vect_force_simple_reduction, vectorizable_reduction): Update
10144 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10146 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
10147 vec<gimple *> to a vec<stmt_vec_info>.
10148 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
10149 the reduction_phis argument from a vec<gimple *> to a
10150 vec<stmt_vec_info>.
10151 (vectorizable_reduction): Likewise the phis local variable that
10152 is passed to vect_create_epilog_for_reduction. Update for new type
10153 of SLP_TREE_VEC_STMTS.
10154 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
10155 (vectorizable_live_operation): Likewise.
10156 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
10157 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
10159 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10161 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
10162 a gimple stmt to a stmt_vec_info.
10163 (vectorizable_condition, vectorizable_live_operation)
10164 (vectorizable_reduction, vectorizable_induction): Pass back the
10165 vectorized statement as a stmt_vec_info.
10166 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
10167 use of STMT_VINFO_VEC_STMT.
10168 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
10169 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
10170 as stmt_vec_infos rather than gimple stmts.
10171 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
10172 to a stmt_vec_info.
10173 (vectorizable_live_operation): Likewise.
10174 (vectorizable_reduction, vectorizable_induction): Likewise,
10175 updating use of STMT_VINFO_VEC_STMT.
10176 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
10177 of STMT_VINFO_VEC_STMT.
10178 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
10179 (vectorizable_simd_clone_call, vectorizable_conversion)
10180 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10181 (vectorizable_store, vectorizable_load, vectorizable_condition)
10182 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
10183 from a gimple stmt to a stmt_vec_info.
10184 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
10185 pointer to a stmt_vec_info to the vectorizable_* routines.
10187 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10189 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
10190 a gimple stmt to a stmt_vec_info.
10191 (is_pattern_stmt_p): Update accordingly.
10192 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
10193 (vect_record_grouped_load_vectors): Likewise.
10194 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
10195 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
10196 (vect_model_reduction_cost): Likewise.
10197 (vect_create_epilog_for_reduction): Likewise.
10198 (vectorizable_reduction, vectorizable_induction): Likewise.
10199 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10200 Return the stmt_vec_info for the pattern statement.
10201 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
10202 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
10203 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
10204 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
10205 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
10206 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
10207 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
10208 (free_stmt_vec_info, vect_is_simple_use): Likewise.
10210 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10212 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
10213 (vect_finish_stmt_generation): Likewise.
10214 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
10215 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
10216 (vect_build_gather_load_calls): Use the return value of the above
10217 functions instead of a separate call to vinfo_for_stmt. Use narrow
10218 scopes for the input gimple stmt and wider scopes for the associated
10219 stmt_vec_info. Use vec_info::lookup_def when setting these
10220 stmt_vec_infos from an SSA_NAME definition.
10221 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10222 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10223 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10224 (vectorizable_store, vectorizable_load, vectorizable_condition)
10225 (vectorizable_comparison): Likewise.
10226 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
10227 (vectorizable_reduction): Likewise.
10229 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10231 * tree-vectorizer.h (vect_is_simple_use): Add an optional
10232 stmt_vec_info * parameter before the optional gimple **.
10233 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10234 (process_use, vect_get_vec_def_for_operand_1): Update callers.
10235 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
10236 * tree-vect-loop.c (vectorizable_reduction): Likewise.
10237 (vectorizable_live_operation): Likewise.
10238 * tree-vect-patterns.c (type_conversion_p): Likewise.
10239 (vect_look_through_possible_promotion): Likewise.
10240 (vect_recog_rotate_pattern): Likewise.
10241 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
10243 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10245 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
10246 a typedef to a wrapper class.
10247 (NULL_STMT_VEC_INFO): New macro.
10248 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
10249 (stmt_vec_info::operator*): New function.
10250 (stmt_vec_info::operator gimple *): Likewise.
10251 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
10252 (add_stmt_costs): Likewise.
10253 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
10254 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
10255 (vect_get_known_peeling_cost): Likewise.
10256 (vect_estimate_min_profitable_iters): Likewise.
10257 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10258 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
10259 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
10260 (vectorizable_store, free_stmt_vec_infos): Likewise.
10261 (new_stmt_vec_info): Change return type of xcalloc to
10264 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10266 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
10267 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
10268 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
10269 a single_imm_use-based sequence.
10270 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
10272 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10274 * tree-vectorizer.h (vec_info::lookup_def): Declare.
10275 * tree-vectorizer.c (vec_info::lookup_def): New function.
10276 * tree-vect-patterns.c (vect_get_internal_def): Use it.
10277 (vect_widened_op_tree): Likewise.
10278 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10279 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
10280 (vectorizable_reduction): Likewise.
10281 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
10283 (vect_is_slp_reduction): Update calls accordingly. Use
10284 vec_info::lookup_def.
10285 (vect_is_simple_reduction): Likewise
10286 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
10288 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10290 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
10291 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
10292 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
10294 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
10295 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
10296 (vect_update_vf_for_slp, vect_analyze_loop_operations)
10297 (vect_is_slp_reduction, vectorizable_induction)
10298 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
10299 * tree-vect-patterns.c (vect_init_pattern_stmt):
10300 (vect_determine_min_output_precision_1, vect_determine_precisions)
10301 (vect_pattern_recog): Likewise.
10302 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
10303 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
10304 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
10305 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
10306 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
10307 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
10308 info field from a loop to a loop_vec_info.
10310 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10312 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
10313 (vec_info::add_stmt): Declare.
10314 * tree-vectorizer.c (vec_info::add_stmt): New function.
10315 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
10316 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
10317 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
10318 (vectorizable_induction): Likewise.
10319 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
10320 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
10321 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
10322 (vectorizable_load): Likewise.
10323 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10324 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
10325 (vect_recog_gather_scatter_pattern): Likewise.
10326 (append_pattern_def_seq): Likewise. Remove a check that is
10327 performed by add_stmt itself.
10329 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10331 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
10332 which make_ssa_name was called with new_stmt before new_stmt
10335 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10337 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
10339 (vect_is_slp_reduction): ...here...
10340 (vect_is_simple_reduction): ...and here. Remove repetition of tests
10341 that are already known to be false.
10343 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10345 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
10346 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
10347 STMT_VINFO_NUM_SLP_USES when it's true.
10348 (vect_free_slp_instance): Add a final_p parameter and pass it to
10349 vect_free_slp_tree.
10350 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
10351 (vect_analyze_slp_instance): Likewise.
10352 (vect_slp_analyze_operations): Likewise.
10353 (vect_slp_analyze_bb_1): Likewise.
10354 * tree-vectorizer.c (vec_info): Likewise.
10355 * tree-vect-loop.c (vect_transform_loop): Likewise.
10357 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10359 * tree-vect-loop.c (vectorizable_reduction): Assert that the
10360 function is not called for second and subsequent members of
10363 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10365 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
10366 cases for nested loops from here to ...
10367 (vect_create_epilog_for_reduction): ...here. Only call
10368 vect_is_simple_use for inner-loop reductions.
10370 2018-07-31 Martin Liska <mliska@suse.cz>
10372 PR gcov-profile/85338
10373 PR gcov-profile/85350
10374 PR gcov-profile/85372
10375 * profile.c (struct location_triplet): New.
10376 (struct location_triplet_hash): Likewise.
10377 (output_location): Do not output a BB that
10378 is already recorded for a line.
10379 (branch_prob): Use streamed_locations.
10381 2018-07-31 Martin Liska <mliska@suse.cz>
10383 PR gcov-profile/85370
10384 * coverage.c (coverage_begin_function): Do not mark target
10385 clones as artificial functions.
10387 2018-07-31 Martin Liska <mliska@suse.cz>
10389 PR gcov-profile/83813
10390 PR gcov-profile/84758
10391 PR gcov-profile/85217
10392 PR gcov-profile/85332
10393 * profile.c (branch_prob): Do not record GOTO expressions
10394 for GIMPLE statements which locations are already streamed.
10396 2018-07-31 Olivier Hainque <hainque@adacore.com>
10398 * gcc.c (handle_spec_function): Accept a soft_matched_part
10399 argument, as do_spec_1. Pass it down to ...
10400 (eval_spec_function): Accept a soft_matched_part argument,
10401 and pass it down to ...
10402 (do_spec_2): Accept a soft_matched_part argument, and pass
10403 it down to do_spec_1.
10404 (do_spec_1): Pass soft_matched_part to handle_spec_function.
10405 (handle_braces): Update call to handle_spec_function.
10406 (driver::set_up_specs): Update calls to do_spec_2.
10407 (compare_debug_dump_opt_spec_function): Likewise.
10408 (compare_debug_self_opt_spec_function): Likewise.
10410 2018-07-31 Olivier Hainque <hainque@adacore.com>
10412 * common.opt (nolibc): New option.
10413 * doc/invoke.texi (Link Options): Document it.
10414 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
10415 * config/alpha/linux.h: Likewise.
10416 * config/arc/elf.h: Likewise.
10417 * config/arm/uclinux-elf.h: Likewise.
10418 * config/arm/unknown-elf.h: Likewise.
10419 * config/avr/avrlibc.h: Likewise.
10420 * config/bfin/bfin.h: Likewise.
10421 * config/bfin/linux.h: Likewise.
10422 * config/bfin/uclinux.h: Likewise.
10423 * config/darwin.h: Likewise.
10424 * config/darwin10.h: Likewise.
10425 * config/darwin12.h: Likewise.
10426 * config/gnu-user.h: Likewise.
10427 * config/lm32/uclinux-elf.h: Likewise.
10428 * config/pa/pa-hpux11.h: Likewise.
10429 * config/pa/pa64-hpux.h: Likewise.
10430 * config/sparc/sparc.h: Likewise.
10432 2018-07-31 Olivier Hainque <hainque@adacore.com>
10434 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
10435 undefined variables.
10437 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
10440 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
10441 instead of GEN_INT.
10443 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
10445 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
10446 terminated string literal.
10448 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
10450 PR rtl-optimization/85160
10451 * combine.c (is_just_move): New function.
10452 (try_combine): Allow combining two instructions into two if neither of
10453 the original instructions was a move.
10455 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
10458 * doc/extend.texi (Global Register Variables): Discourage use of type
10460 (Local Register Variables): Likewise.
10462 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
10464 PR tree-optimization/86506
10465 * hwint.h (ceil_log2): Resync with hwint.c implementation.
10467 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
10470 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
10471 hard_regno, make sure no insn between `from` and `to` clobbers it.
10473 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
10474 Tom de Vries <tdevries@suse.de>
10476 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
10477 (PTX_DEFAULT_RUNTIME_DIM): ... this.
10478 (nvptx_goacc_validate_dims): Set default worker and gang dims to
10479 PTX_DEFAULT_RUNTIME_DIM.
10480 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
10482 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
10484 * config/pa/pa.c (pa_output_addr_vec): Align address table.
10485 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
10486 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
10488 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
10490 * config/rs6000/constraints.md (wG constraint): Delete, no longer
10492 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
10493 predicate to reflect toc fusion has been deleted.
10494 (toc_fusion_mem_raw): Delete, no longer used.
10495 (toc_fusion_mem_wrapped): Likewise.
10496 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
10498 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
10499 Delete, no longer used.
10500 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
10501 meant to be used for toc fusion.
10502 (rs6000_debug_print_mode): Delete toc fusion debugging.
10503 (rs6000_debug_reg_global): Likewise.
10504 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
10505 fusion and secondary reload support that were never used.
10506 (rs6000_option_override_internal): Delete TOC fusion, that was only
10507 partially defined, and it did not work unless you also used the
10509 (rs6000_legitimate_address_p): Delete TOC fusion support.
10510 (rs6000_opt_masks): Likewise.
10511 (fusion_wrap_memory_address): Delete function, no longer used.
10512 (fusion_split_address); Delete TOC fusion support.
10513 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
10514 longer used with toc fusion being deleted.
10515 (TARGET_TOC_FUSION_FP): Likewise.
10516 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
10518 (toc fusion spliter): Delete TOC fusion support.
10519 (toc_fusionload_<mode>): Likewise.
10520 (toc_fusionload_di): Likewise.
10521 (fusion_gpr_load_<mode>): Delete generator function, this insn no
10522 longer needs to be named. Rename predicate to delete TOC fusion.
10523 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
10524 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
10525 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
10526 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
10527 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
10529 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
10531 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
10532 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
10533 __int128 in built-in function prototypes.
10534 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
10535 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
10537 2018-07-27 Martin Sebor <msebor@redhat.com>
10539 PR tree-optimization/86696
10540 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
10541 types, including enums.
10542 (handle_char_store): Be prepared for the above function to fail.
10544 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
10546 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
10547 when optimization level is lower than 2 or optimize for size.
10549 2018-07-26 Martin Sebor <msebor@redhat.com>
10551 PR tree-optimization/86043
10552 PR tree-optimization/86042
10553 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
10554 (get_string_cst_length): Rename...
10555 (get_min_string_length): ...to this. Add argument.
10556 (handle_char_store): Extend to handle multi-character stores by
10558 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
10559 * tree.h (initializer_zerop): Add argument.
10561 2018-07-26 Jakub Jelinek <jakub@redhat.com>
10563 PR middle-end/86660
10564 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
10565 declare target to variables if they have always,{to,from,tofrom} map
10568 2018-07-26 Martin Liska <mliska@suse.cz>
10571 * lto-wrapper.c: Add linker_output as prefix
10572 for ltrans_output_file.
10574 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
10576 PR rtl-optimization/85805
10577 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
10578 value for hard registers if that was written in the same mode.
10580 2018-07-26 Martin Liska <mliska@suse.cz>
10582 PR gcov-profile/86536
10583 * gcov.c (format_gcov): Use printf format %.*f directly
10584 and do not handle special values.
10586 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
10588 * common/config/arc/arc-common.c (arc_option_optimization_table):
10589 Update default optimizations for size.
10591 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
10593 * config/arc/arc.md (movsf_insn): Add short instruction selection.
10594 * config/arc/constraints.md (CfZ): New constraint.
10595 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
10596 (subsf3_fpu): Likewise.
10597 (cmpsf_fpu): Likewise.
10598 (cmpsf_fpu_uneq): Likewise.
10600 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
10602 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
10604 (arc_is_uncached_mem_p): uncached applies to both the variable and
10607 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
10609 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
10612 2018-07-25 David Malcolm <dmalcolm@redhat.com>
10614 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
10615 field "m_scopes" from vec to auto_vec.
10617 2018-07-25 Martin Liska <mliska@suse.cz>
10619 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
10622 2018-07-25 Richard Biener <rguenther@suse.de>
10625 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
10626 special wrt context_die late.
10627 (gen_subprogram_die): Re-use DIEs in local scope.
10629 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
10631 PR tree-optimization/86644
10632 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
10634 2018-07-25 Martin Liska <mliska@suse.cz>
10636 PR middle-end/86645
10637 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
10638 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
10640 2018-07-25 Martin Liska <mliska@suse.cz>
10643 * params.def: Explain ASan abbreviation and provide
10644 a documentation link.
10646 2018-07-24 Martin Sebor <msebor@redhat.com>
10648 PR tree-optimization/86622
10649 PR tree-optimization/86532
10650 * builtins.h (string_length): Declare.
10651 * builtins.c (c_strlen): Correct handling of non-constant offsets.
10652 (check_access): Be prepared for non-constant length ranges.
10653 (string_length): Make extern.
10654 * expr.c (string_constant): Only handle the minor non-constant
10655 array index. Use string_constant to compute the length of
10656 a generic string constant.
10658 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
10660 PR tree-optimization/86618
10661 * tree-vect-stmts.c (vectorizable_call): Don't take the address
10662 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
10664 2018-07-24 David Malcolm <dmalcolm@redhat.com>
10666 PR tree-optimization/86636
10667 * json.cc (json::object::set): Fix comment. Add assertions.
10668 (json::array::append): Move here from json.h. Add comment and an
10670 (json::string::string): Likewise.
10671 * json.h (json::array::append): Move to json.cc.
10672 (json::string::string): Likewise.
10673 * optinfo-emit-json.cc
10674 (optrecord_json_writer::impl_location_to_json): Assert that we
10675 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
10676 wrapper around it. Expand the location once, rather than three
10678 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
10679 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
10681 (optrecord_json_writer::optinfo_to_json): Likewise, in four
10682 places. Fix some overlong lines.
10684 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
10686 * config/aarch64/aarch64-simd.md
10687 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
10688 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
10689 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
10690 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
10691 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
10692 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
10693 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
10694 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
10695 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
10697 2018-07-24 Jakub Jelinek <jakub@redhat.com>
10699 PR middle-end/86627
10700 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
10701 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
10702 and abs_d == d, do the power of two handling if profitable.
10704 2018-07-24 Richard Biener <rguenther@suse.de>
10706 * match.pd: Add BIT_FIELD_REF canonicalizations.
10708 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
10711 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
10713 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
10715 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
10716 terminated STRING_CST object.
10718 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
10720 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
10722 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
10724 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
10725 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
10726 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
10727 the elements into a register.
10728 (rs6000_split_v4si_init_di_reg): Delete.
10729 (rs6000_split_v4si_init): Delete.
10730 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
10731 (vsx_init_v4si): Rewrite as a define_expand.
10733 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
10735 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
10736 zero_extend argument from memory): New.
10738 2018-07-22 Martin Sebor <msebor@redhat.com>
10741 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
10742 diagnosing calls with unknown arguments unless -Walloca-larger-than
10743 is restricted to less than PTRDIFF_MAX bytes.
10745 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
10747 * doc/gcov.texi (Invoking Gcov): Editorial changes.
10749 2018-07-20 David Malcolm <dmalcolm@redhat.com>
10751 * pretty-print.c (text_info::set_location): Remove redundant
10752 "line_table" parameter from call to rich_location::set_range.
10754 2018-07-20 Martin Sebor <msebor@redhat.com>
10756 PR middle-end/82063
10757 * builtins.c (expand_builtin_alloca): Adjust.
10758 * calls.c (alloc_max_size): Simplify.
10759 * cgraphunit.c (cgraph_node::expand): Adjust.
10760 * common.opt (larger_than_size, warn_frame_larger_than): Remove
10762 (frame_larger_than_size): Same.
10763 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
10764 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
10766 * doc/invoke.texi (GCC Command Options): Document option arguments.
10767 Explain byte-size arguments and suffixes.
10768 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
10769 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
10770 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
10771 * doc/options.texi (UInteger): Expand.
10772 (Host_Wide_Int, ByteSize): Document new properties.
10773 * final.c (final_start_function_1): Include sizes in an error message.
10774 * function.c (frame_offset_overflow): Same.
10775 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
10776 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
10777 Diagnose unbounded alloca calls only for limits of less than
10779 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
10780 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
10782 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
10783 only for limits of less than PTRDIFF_MAX.
10784 * langhooks-def.h (lhd_handle_option): Change function argument
10786 * langhooks.c (lhd_handle_option): Same.
10787 * langhooks.h (handle_option): Same.
10788 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
10790 (var_type, var_type_struct): Same.
10791 (var_set): Handle ByteSize flag.
10792 * optc-gen.awk: Add comments to output to ease debugging. Make
10793 use of HOST_WIDE_INT where appropriate.
10794 * opts-gen-save.awk: Use %lx to format unsigned long.
10795 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
10796 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
10797 arguments. Parse bytes-size suffixes.
10798 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
10799 (enum_value_to_arg): Same.
10800 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
10801 (handle_option): Adjust.
10802 (generate_option): Change function argument to HOST_WIDE_INT.
10803 (cmdline_handle_error): Adjust.
10804 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
10805 (set_option): Change function argument to HOST_WIDE_INT.
10806 (option_enabled): Handle cl_host_wide_int.
10807 (get_option_state): Handle CLVC_SIZE.
10808 (control_warning_option): Same.
10809 * opts.c (common_handle_option): Change function argument to
10810 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
10811 OPT_Wvla_larger_than_.
10812 * opts.h (enum cl_var_type): Add an enumerator.
10813 * stor-layout.c (layout_decl): Print a more meaningful warning.
10814 * toplev.c (output_stack_usage): Adjust.
10816 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
10818 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
10819 call to inline_expand_builtin_string_cmp.
10820 (expand_builtin_strcmp): Likewise.
10821 (expand_builtin_strncmp): Likewise.
10822 (inline_string_cmp): Delete the last parameter, change char_type_node
10823 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
10825 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
10826 the inlining expansion on target where the type of the call has same or
10827 narrower precision than unsigned char.
10829 2018-07-20 David Malcolm <dmalcolm@redhat.com>
10831 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
10832 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
10833 * common.opt (fsave-optimization-record): New option.
10834 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
10835 * doc/invoke.texi (-fsave-optimization-record): New option.
10836 * dumpfile.c: Include "optinfo-emit-json.h".
10837 (struct kv_pair): Move to coretypes.h.
10838 (optgroup_options): Make non-static.
10839 (dump_context::end_scope): Call
10840 optimization_records_maybe_pop_dump_scope.
10841 * dumpfile.h (optgroup_options): New decl.
10842 * json.cc: New file.
10843 * json.h: New file.
10844 * optinfo-emit-json.cc: New file.
10845 * optinfo-emit-json.h: New file.
10846 * optinfo.cc: Include "optinfo-emit-json.h".
10847 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
10848 (optinfo_enabled_p): Check optimization_records_enabled_p.
10849 (optinfo_wants_inlining_info_p): Likewise.
10850 * optinfo.h: Update comment.
10851 * profile-count.c (profile_quality_as_string): New function.
10852 * profile-count.h (profile_quality_as_string): New decl.
10853 (profile_count::quality): New accessor.
10854 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
10855 and optinfo_emit_json_cc_tests.
10856 * selftest.h (selftest::json_cc_tests): New decl.
10857 (selftest::optinfo_emit_json_cc_tests): New decl.
10858 * toplev.c: Include "optinfo-emit-json.h".
10859 (compile_file): Call optimization_records_finish.
10860 (do_compile): Call optimization_records_start.
10861 * tree-ssa-live.c: Include optinfo.h.
10862 (remove_unused_scope_block_p): Retain inlining information if
10863 optinfo_wants_inlining_info_p returns true.
10865 2018-07-20 Richard Biener <rguenther@suse.de>
10868 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
10869 to cover -flto-partition=none.
10871 2018-07-20 Martin Liska <mliska@suse.cz>
10873 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
10874 (get_decl_source_range): Remove unused function.
10876 2018-07-20 Richard Biener <rguenther@suse.de>
10878 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
10879 (struct vn_phi_s): Likewise.
10880 (struct vn_reference_s): Likewise.
10881 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
10882 for searching the slot of an entry known to be in the hash itself.
10883 (vn_phi_hasher::equal): Likewise.
10884 (vn_reference_hasher::equal): Likewise.
10885 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
10887 (optimistic_info, current_info): Remove, keeping only valid_info.
10888 (vn_reference_lookup_1): Remove fallback lookup.
10889 (vn_reference_lookup_2): Likewise.
10890 (vn_nary_op_lookup_1): Likewise.
10891 (vn_phi_lookup): Likewise.
10892 (vn_nary_build_or_lookup_1): Make sure to not chain the built
10894 (vn_reference_insert): Adjust, chain the inserted hash element
10895 at last_inserted_ref.
10896 (vn_reference_insert_pieces): Likewise.
10897 (visit_reference_op_call): Likewise.
10898 (vn_nary_op_insert_into): Chain the inserted hash element at
10899 last_inserted_nary.
10900 (vn_nary_op_insert_pieces): Adjust.
10901 (vn_nary_op_insert): Likewise.
10902 (vn_nary_op_insert_stmt): Likewise.
10903 (vn_phi_insert): Adjust, chain the inserted hash element at
10905 (process_scc): Remove clearing and copying the optimistic
10906 table. Instead remove elements inserted during an optimistic
10907 iteration from the single table we maintain.
10908 (init_scc_vn): Adjust.
10909 (free_scc_vn): Likewise.
10910 (sccvn_dom_walker::record_cond): Likewise.
10911 (sccvn_dom_walker::after_dom_children): Likewise.
10913 2018-07-19 Martin Sebor <msebor@redhat.com>
10915 PR tree-optimization/84047
10916 PR tree-optimization/83776
10917 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
10918 (check_array_bounds): Call it.
10920 2018-07-19 Martin Sebor <msebor@redhat.com>
10922 * align.h (align_flags): Use member initialization.
10924 2018-07-19 David Malcolm <dmalcolm@redhat.com>
10926 * Makefile.in (OBJS): Add optinfo.o.
10927 * coretypes.h (class symtab_node): New forward decl.
10928 (struct cgraph_node): New forward decl.
10929 (class varpool_node): New forward decl.
10930 * dump-context.h: New file.
10931 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
10933 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
10934 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
10935 (set_alt_dump_file): Likewise.
10936 (dump_context::~dump_context): New dtor.
10937 (dump_gimple_stmt): Move implementation to...
10938 (dump_context::dump_gimple_stmt): ...this new member function.
10939 Add the stmt to any pending optinfo, creating one if need be.
10940 (dump_gimple_stmt_loc): Move implementation to...
10941 (dump_context::dump_gimple_stmt_loc): ...this new member function.
10942 Start a new optinfo and add the stmt to it.
10943 (dump_gimple_expr): Move implementation to...
10944 (dump_context::dump_gimple_expr): ...this new member function.
10945 Add the stmt to any pending optinfo, creating one if need be.
10946 (dump_gimple_expr_loc): Move implementation to...
10947 (dump_context::dump_gimple_expr_loc): ...this new member function.
10948 Start a new optinfo and add the stmt to it.
10949 (dump_generic_expr): Move implementation to...
10950 (dump_context::dump_generic_expr): ...this new member function.
10951 Add the tree to any pending optinfo, creating one if need be.
10952 (dump_generic_expr_loc): Move implementation to...
10953 (dump_context::dump_generic_expr_loc): ...this new member
10954 function. Add the tree to any pending optinfo, creating one if
10956 (dump_printf): Move implementation to...
10957 (dump_context::dump_printf_va): ...this new member function. Add
10958 the text to any pending optinfo, creating one if need be.
10959 (dump_printf_loc): Move implementation to...
10960 (dump_context::dump_printf_loc_va): ...this new member function.
10961 Start a new optinfo and add the stmt to it.
10962 (dump_dec): Move implementation to...
10963 (dump_context::dump_dec): ...this new member function. Add the
10964 value to any pending optinfo, creating one if need be.
10965 (dump_context::dump_symtab_node): New member function.
10966 (dump_context::get_scope_depth): New member function.
10967 (dump_context::begin_scope): New member function.
10968 (dump_context::end_scope): New member function.
10969 (dump_context::ensure_pending_optinfo): New member function.
10970 (dump_context::begin_next_optinfo): New member function.
10971 (dump_context::end_any_optinfo): New member function.
10972 (dump_context::s_current): New global.
10973 (dump_context::s_default): New global.
10974 (dump_scope_depth): Delete global.
10975 (dumpfile_ensure_any_optinfo_are_flushed): New function.
10976 (dump_symtab_node): New function.
10977 (get_dump_scope_depth): Reimplement in terms of dump_context.
10978 (dump_begin_scope): Likewise.
10979 (dump_end_scope): Likewise.
10980 (selftest::temp_dump_context::temp_dump_context): New ctor.
10981 (selftest::temp_dump_context::~temp_dump_context): New dtor.
10982 (selftest::verify_item): New function.
10983 (ASSERT_IS_TEXT): New macro.
10984 (ASSERT_IS_TREE): New macro.
10985 (ASSERT_IS_GIMPLE): New macro.
10986 (selftest::test_capture_of_dump_calls): New test.
10987 (selftest::dumpfile_c_tests): Call it.
10988 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
10989 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
10990 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
10991 descriptive comment.
10992 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
10993 (dump_node, dump_bb): Move these unrelated decls.
10994 (class dump_manager): Add leading comment.
10995 * optinfo.cc: New file.
10996 * optinfo.h: New file.
10998 2018-07-19 Michael Collison <michael.collison@arm.com>
10999 Richard Henderson <rth@redhat.com>
11001 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
11002 (subti): Handle op1 zero.
11003 (subvti4, usub4ti4): New.
11004 (*sub<GPI>3_compare1_imm): New.
11005 (sub<GPI>3_carryinCV): New.
11006 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
11007 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
11009 2018-07-19 Michael Collison <michael.collison@arm.com>
11010 Richard Henderson <rth@redhat.com>
11012 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
11013 (addti3): Create simpler code if low part is already known to be 0.
11014 (addvti4, uaddvti4): New.
11015 (*add<GPI>3_compareC_cconly_imm): New.
11016 (*add<GPI>3_compareC_cconly): New.
11017 (*add<GPI>3_compareC_imm): New.
11018 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
11019 handle constants within this pattern..
11020 (*add<GPI>3_compareV_cconly_imm): New.
11021 (*add<GPI>3_compareV_cconly): New.
11022 (*add<GPI>3_compareV_imm): New.
11023 (add<GPI>3_compareV): New.
11024 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
11025 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
11026 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
11027 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
11028 with 'comparison' operator.
11029 (*add<GPI>3_compareV_cconly_imm): Ditto.
11030 (*add<GPI>3_compareV_cconly): Ditto.
11031 (*add<GPI>3_compareV_imm): Ditto.
11032 (add<GPI>3_compareV): Ditto.
11033 (add<mode>3_carryinC): Ditto.
11034 (*add<mode>3_carryinC_zero): Ditto.
11035 (*add<mode>3_carryinC): Ditto.
11036 (add<mode>3_carryinV): Ditto.
11037 (*add<mode>3_carryinV_zero): Ditto.
11038 (*add<mode>3_carryinV): Ditto.
11040 2018-07-19 Michael Collison <michael.collison@arm.com>
11041 Richard Henderson <rth@redhat.com>
11043 * config/aarch64/aarch64-modes.def (CC_V): New.
11044 * config/aarch64/aarch64-protos.h
11045 (aarch64_addti_scratch_regs): Declare
11046 (aarch64_subvti_scratch_regs): Declare.
11047 (aarch64_expand_subvti): Declare.
11048 (aarch64_gen_unlikely_cbranch): Declare
11049 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
11050 for signed overflow using CC_Vmode.
11051 (aarch64_get_condition_code_1): Handle CC_Vmode.
11052 (aarch64_gen_unlikely_cbranch): New function.
11053 (aarch64_addti_scratch_regs): New function.
11054 (aarch64_subvti_scratch_regs): New function.
11055 (aarch64_expand_subvti): New function.
11057 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
11059 * config/aarch64/aarch64-option-extensions.def: New entry for profile
11061 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
11062 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
11065 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
11068 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11069 address check not strict.
11071 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
11073 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
11075 (store_pair_lanes<mode>): Likewise.
11076 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
11077 enum value 'ADDR_QUERY_LDP_STP_N'.
11078 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
11079 (aarch64_print_address_internal): Add declaration.
11080 (aarch64_print_ldpstp_address): Remove.
11081 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
11082 (aarch64_print_operand): Change printing of 'y'.
11083 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
11084 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
11085 'true' rather than '1'.
11086 * config/aarch64/constraints.md (Uml): Likewise.
11087 (Uml): Rename to Umn.
11090 2018-07-19 Richard Biener <rguenther@suse.de>
11092 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
11094 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
11095 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
11096 (vn_reference_hasher): Likewise.
11097 (struct vn_tables_s): Remove obstack and alloc-pool members.
11098 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
11099 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
11100 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
11101 (vn_reference_insert_pieces): Likewise.
11102 (alloc_vn_nary_op_noinit): Adjust.
11103 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
11104 (vn_phi_eq): Adjust.
11105 (shared_lookup_phiargs): Remove.
11106 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
11107 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
11108 (visit_reference_op_call): Likewise.
11109 (copy_nary, copy_phi, copy_reference): Remove.
11110 (process_scc): Rewind the obstack when iterating. Do not
11111 copy the elements to valid_info but just move them from one
11112 hashtable to the other.
11113 (allocate_vn_table): Adjust.
11114 (free_vn_table): Likewise.
11115 (init_scc_vn): Likewise.
11116 (free_scc_vn): Likewise.
11118 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
11121 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
11122 indirect_return as function type attribute.
11123 (ix86_attribute_table): Change indirect_return to function
11125 * doc/extend.texi: Update indirect_return attribute.
11127 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
11129 * wide-int.h (widest2_int): New.
11130 * gimple-fold.c (arith_overflowed_p): Use it.
11131 * tree.h (widest2_int_cst): New.
11132 * tree-vrp.c (wide_int_binop_overflow): Rename from
11133 vrp_int_const_binop.
11134 Rewrite to work on trees.
11135 (extract_range_from_multiplicative_op_1): Abstract code to...
11136 (wide_int_range_min_max): ...here.
11137 (wide_int_range_cross_product): ...and here.
11138 (extract_range_from_binary_expr_1): Abstract overflow code to...
11139 (wide_int_range_mult_wrapping): ...here.
11140 * tree-vrp.h (wide_int_range_cross_product): New.
11141 (wide_int_range_mult_wrapping): New.
11143 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
11144 Julia Koval <julia.koval@intel.com>
11146 * config/i386/x86-tune-costs.h (skylake_memcpy,
11147 skylake_memset): Replace rep_prefix with unrolling for size 512.
11149 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
11151 PR middle-end/86544
11152 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
11153 comparision with EQ_EXPR in last stmt.
11155 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
11157 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
11158 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
11159 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
11160 previously known as "PowerPC AltiVec Built-in Functions". Move
11161 some material to new subsubsections "PowerPC AltiVec Built-in
11162 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
11164 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
11165 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
11166 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
11167 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
11169 2018-07-18 Richard Biener <rguenther@suse.de>
11171 PR tree-optimization/86557
11172 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
11175 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
11177 * config/s390/s390.c (s390_function_profiler): Generate CFI.
11179 2018-07-17 Jeff Law <law@redhat.com>
11181 * config/arm/arm.c (get_label_padding): Update for recent
11182 changes to label_to_alignment.
11184 PR tree-optimization/86010
11185 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
11187 * config/mips/mips.c (vr4130_align_insns): Update for recent
11188 changes to label_to_alignment.
11190 * config/frv/frv.c (frv_label_align): Update for recent changes
11191 to label_to_alignment.
11193 * config/nios2/nios2.c (nios2_label_align): Update for recent
11194 changes which dropped ALIGN_LABELS_LOG.
11196 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
11198 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
11199 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
11201 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
11203 * config/arc/arc.c (arc_label_align): Use align_labels instead of
11204 deprecated align_labels_log.
11206 2018-07-17 Richard Biener <rguenther@suse.de>
11209 * dwarf2out.c (init_sections_and_labels): Always generate
11210 a debug_line_str_section for early LTO debug.
11211 (dwarf2out_finish): Reset debug_line_str_hash output early.
11212 Bump counter for extra dwarf5 .debug_loc labels to not conflict
11214 (dwarf2out_early_finish): Output debug_line_str.
11216 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
11218 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
11219 index register on z196 or later.
11221 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
11223 * config/s390/s390.c (s390_default_align): Set default function
11225 (s390_override_options_after_change): Call s390_default align.
11226 (s390_option_override_internal): Call s390_default_align.
11227 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
11229 2018-07-17 Jakub Jelinek <jakub@redhat.com>
11231 PR middle-end/86542
11232 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
11233 to _looptemp_ clauses, other than the first two.
11235 2018-07-17 Martin Liska <mliska@suse.cz>
11237 * opts.c: Do not enable OPT_falign_* for -Os.
11239 2018-07-17 Martin Liska <mliska@suse.cz>
11241 * align.h (MAX_CODE_ALIGN): New.
11242 (MAX_CODE_ALIGN_VALUE): New.
11243 * common/config/i386/i386-common.c (ix86_handle_option):
11244 (MAX_CODE_ALIGN): Moved to align.h.
11245 * final.c (MAX_CODE_ALIGN): Likewise.
11246 * opts.c (parse_and_check_align_values):
11247 (MAX_CODE_ALIGN): Likewise.
11248 (MAX_CODE_ALIGN_VALUE): Likewise.
11250 2018-07-17 Martin Liska <mliska@suse.cz>
11252 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
11253 in order to fulfil coding style.
11254 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
11255 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11256 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11257 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
11258 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
11259 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
11260 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
11261 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
11262 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11264 2018-07-17 Martin Liska <mliska@suse.cz>
11266 * align.h: New file.
11267 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
11269 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
11270 align_flags of label_to_alignment.
11271 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
11273 * config/m68k/m68k.c: Do not use removed align_labels_value and
11275 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
11276 (LOOP_ALIGN): Likewise.
11277 (LABEL_ALIGN): Likewise.
11278 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
11279 Remove not used macro.
11280 (rs6000_loop_align): Change return type to align_flags.
11281 (rs6000_loop_align_max_skip): Remove.
11282 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
11283 Change return type to align_flags.
11284 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
11285 Remove not used macro.
11286 (rs6000_loop_align): Change return type to align_flags.
11287 (rs6000_loop_align_max_skip): Remove.
11288 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
11289 * config/rx/rx-protos.h (rx_align_for_label): Make it
11291 * config/rx/rx.c (rx_align_for_label): Change return type
11293 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
11295 into align_flags class.
11296 (LABEL_ALIGN): Likewise.
11297 (LOOP_ALIGN): Likewise.
11298 * config/s390/s390.c (s390_label_align): Use align_flags
11300 (s390_asm_output_function_label): Likewise.
11301 * config/sh/sh.c (sh_override_options_after_change):
11302 Use align_flags class directly without macros.
11303 (find_barrier): Likewise.
11304 (barrier_align): Likewise.
11305 (sh_loop_align): Likewise.
11306 * config/spu/spu.c (spu_option_override):
11307 Use align_flags_tuple::get_value instead of removed macros.
11308 (spu_sched_init): Likewise.
11309 * config/spu/spu.h (GTY): Likewise.
11310 * config/visium/visium.c (visium_option_override):
11311 Set "8" as default secondary alignment.
11312 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
11313 in order to guarantee secondary alignment of 8.
11314 * coretypes.h: Include align.h header file.
11315 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
11316 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
11317 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
11318 * doc/tm.texi.in: Likewise.
11319 * final.c (struct label_alignment): Remove not used structure.
11320 (LABEL_ALIGN): Change type to align_flags.
11321 (LOOP_ALIGN): Likewise.
11322 (JUMP_ALIGN): Likewise.
11323 (default_loop_align_max_skip): Remove.
11324 (default_label_align_max_skip): Likewise.
11325 (default_jump_align_max_skip): Likewise.
11326 (default_label_align_after_barrier_max_skip):
11327 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
11328 (LABEL_TO_MAX_SKIP): Remove.
11329 (label_to_alignment): Return align_flags type instead of integer.
11330 (label_to_max_skip): Remove.
11331 (align_fuzz): Use align_flags type.
11332 (compute_alignments): Use align_flags type and use align_flags::max
11333 to combine multiple alignments.
11334 (grow_label_align): Grow vec instead of C array.
11335 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
11336 (shorten_branches): Use align_flags type and use align_flags::max
11337 to combine multiple alignments.
11338 (final_scan_insn_1): Remove usage of secondary alignment that comes
11339 from label alignment, but instead use proper secondary alignment
11340 which is computed in grow_label_align.
11341 * flags.h (struct align_flags_tuple): Move to align.h.
11342 (struct align_flags): Likewise.
11343 (state_align_loops): Rename to align_loops.
11344 (state_align_jumps): Rename to align_jumps.
11345 (state_align_labels): Rename to align_labels.
11346 (state_align_functions): Rename to align_functions.
11347 (align_loops_log): Remove.
11348 (align_jumps_log): Remove.
11349 (align_labels_log): Remove.
11350 (align_functions_log): Remove.
11351 (align_loops_max_skip): Remove.
11352 (align_jumps_max_skip): Remove.
11353 (align_labels_max_skip): Remove.
11354 (align_functions_max_skip): Remove.
11355 (align_loops_value): Remove.
11356 (align_jumps_value): Remove.
11357 (align_labels_value): Remove.
11358 (align_functions_value): Remove.
11359 * output.h (label_to_alignment): Change return type to align_flags.
11360 (label_to_max_skip): Remove.
11361 * target.def: Remove loop_align_max_skip, label_align_max_skip,
11362 jump_align_max_skip macros.
11363 * targhooks.h (default_loop_align_max_skip): Remove.
11364 (default_label_align_max_skip): Likewise.
11365 (default_jump_align_max_skip): Likewise.
11366 (default_label_align_after_barrier_max_skip): Remove.
11367 * toplev.c (read_log_maxskip): Use ::normalize function.
11368 (parse_N_M): Remove not used argument and also call ::normalize.
11369 (parse_alignment_opts): Do not pass unused arguments.
11370 * varasm.c (assemble_start_function): Use directly align_functions
11371 instead of removed macros.
11372 * system.h: Do not poison removed macros.
11374 2018-07-17 Jakub Jelinek <jakub@redhat.com>
11376 PR middle-end/86539
11377 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
11378 and cond temporaries don't have reference type if iterator has
11379 pointer type. For init use &for_pre_body instead of pre_p if
11380 for_pre_body is non-empty.
11382 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
11384 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
11385 double-double modes to SFmode directly directly.
11386 (trunc<mode>sf2_fprs): Delete.
11388 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
11390 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
11391 for conversions between IFmode and the decimal floating point modes.
11392 (init_float128_ieee): Use the correct names for conversions between
11393 KFmode and the decimal floating point modes.
11395 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
11397 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
11398 for the conversions between TDmode and IFmode.
11399 (init_float128_ieee): Use more correct names for the conversions
11400 between TDmode and KFmode.
11402 2018-07-16 Jakub Jelinek <jakub@redhat.com>
11404 PR tree-optimization/86526
11405 * builtins.c (expand_builtin_memcmp): Formatting fixes.
11406 (inline_expand_builtin_string_cmp): Likewise.
11407 (inline_string_cmp): Likewise. Use c_readstr instead of
11408 builtin_memcpy_read_str. Add unit_mode temporary.
11410 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
11412 PR middle-end/86528
11413 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
11414 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
11416 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
11418 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
11419 Alphabetize prototypes of built-in functions, separating out
11420 built-in functions that are listed in this section but should be
11421 described elsewhere.
11423 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
11426 * expmed.c (emit_store_flag): Do not emit setcc followed by a
11427 conditional move when trapping comparison was split to a
11428 non-trapping one (and vice versa).
11430 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
11432 * config/s390/s390.c (s390_function_profiler): Generate nops
11433 instead of profiler call sequences.
11434 * config/s390/s390.opt: Add the new option.
11436 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
11438 * config/s390/s390.c (s390_function_profiler): Generate
11439 __mcount_loc section.
11440 * config/s390/s390.opt: Add the new option.
11442 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
11444 * common.opt: Add the new warning.
11445 * config/s390/s390.c (s390_function_profiler): Emit "brasl
11446 %r0,__fentry__" when -mfentry is specified.
11447 (s390_option_override_internal): Disallow -mfentry for 31-bit
11449 * config/s390/s390.opt: Add the new option.
11451 2018-07-16 Richard Biener <rguenther@suse.de>
11454 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
11455 for function-local FUNCTION_DECL and RESULT_DECL immediately.
11457 2018-07-16 Martin Liska <mliska@suse.cz>
11460 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
11463 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
11465 * config/arc/arcHS.md: Update ARCHS scheduling rules.
11467 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
11469 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
11471 * config/arc/arc-cpus.def (hs4x): New cpu.
11473 * config/arc/arc-tables.opt: Regenerate.
11474 * config/arc/arc.c (arc_sched_issue_rate): New function.
11475 (TARGET_SCHED_ISSUE_RATE): Define.
11476 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
11477 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
11479 (attr tune): Add ARCHS4x tune values.
11480 (attr tune_dspmpy): Define.
11481 (*tst): Correct instruction type.
11482 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
11483 * config/arc/arcHS4x.md: New file.
11484 * config/arc/fpu.md: Update instruction type attributes.
11485 * config/arc/t-multilib: Regenerate.
11487 2018-07-16 Tom de Vries <tdevries@suse.de>
11490 * var-tracking.c (vt_initialize): Fix pre_dec handling.
11492 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
11494 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
11497 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
11499 PR tree-optimization/86514
11500 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
11501 conversion to a boolean type from a type with greater precision.
11503 2018-07-16 Tom de Vries <tdevries@suse.de>
11505 * var-tracking.c (vt_initialize): Print adjusted insn slim if
11506 dump_flags request TDF_SLIM.
11508 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
11510 * fold-const.c (int_const_binop_1): Abstract...
11511 (wide_int_binop): ...wide int code here.
11512 (poly_int_binop): ...poly int code here.
11513 Abstract the rest of int_const_binop_1 into int_const_binop.
11514 * fold-const.h (wide_int_binop): New.
11515 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
11516 Remove useless PLUS/MINUS_EXPR case.
11517 (zero_nonzero_bits_from_vr): Move wide int code...
11518 (zero_nonzero_bits_from_bounds): ...here.
11519 (extract_range_from_binary_expr_1): Move mask optimization code...
11520 (range_easy_mask_min_max): ...here.
11521 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
11522 (range_easy_mask_min_max): New.
11524 2018-07-15 Jeff Law <law@redhat.com>
11527 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
11530 2018-07-14 Jim Wilson <jimw@sifive.com>
11532 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
11534 2018-07-14 Paul Koning <ni1d@arrl.net>
11536 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
11538 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
11540 * lto-streamer-out.c (copy_function_or_variable): Dump info about
11543 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
11544 Steve Munroe <munroesj52@gmail.com>
11546 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
11547 (_mm_andnot_si128): Likewise.
11548 (_mm_or_si128): Likewise.
11549 (_mm_xor_si128): Likewise.
11551 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
11553 PR middle-end/78809
11554 * builtins.c (expand_builtin_memcmp): Inline the calls first
11555 when result_eq is false.
11556 (expand_builtin_strcmp): Inline the calls first.
11557 (expand_builtin_strncmp): Likewise.
11558 (inline_string_cmp): New routine. Expand a string compare
11559 call by using a sequence of char comparison.
11560 (inline_expand_builtin_string_cmp): New routine. Inline expansion
11561 a call to str(n)cmp/memcmp.
11562 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
11564 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
11566 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
11568 * config/arm/driver-arm.c: Include arm-native.h.
11569 (host_detect_local_cpu): Use auto-generated data tables.
11570 (vendors, arm_cpu_table): Delete. Move part information to ...
11571 * config/arm/arm-cpus.in: ... here.
11572 * config/arm/parsecpu.awk (gen_native): New function.
11573 (vendor, part): New CPU fields.
11574 (END): Add support for building the native CPU detection tables.
11575 * config/arm/t-arm (arm-native.h): Add build rule.
11576 (driver-arm.o): Add dependency on arm-native.h.
11578 2018-07-13 Richard Biener <rguenther@suse.de>
11580 PR middle-end/85974
11581 * match.pd (addr1 - addr2): Allow either of the operand to
11584 2018-07-13 Tom de Vries <tdevries@suse.de>
11586 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
11589 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
11591 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
11592 arrays instead of numbered variables.
11594 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
11596 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
11597 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
11599 2018-07-13 Richard Biener <rguenther@suse.de>
11602 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
11603 instead of get_context_die.
11605 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
11606 Richard Biener <rguenther@suse.de>
11608 PR middle-end/86489
11609 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
11610 that the loop latch destination where phi is defined.
11612 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
11614 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
11615 (riscv_expand_epilogue): Add assertion to check interrupt mode.
11616 (riscv_set_current_function): Extract getting interrupt type to new
11618 (riscv_get_interrupt_type): New function.
11619 (riscv_merge_decl_attributes): New function, checking interrupt type is
11621 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
11623 2018-07-12 Paul Koning <ni1d@arrl.net>
11625 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
11628 2018-07-12 Paul Koning <ni1d@arrl.net>
11630 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
11631 zero reference, add doloop_end instead.
11632 * doc/md.texi (decrement_and_branch_until_zero): Remove.
11633 (Looping patterns): Remove decrement_and_branch_until_zero. Add
11634 detail for doloop_end.
11636 2018-07-12 Martin Sebor <msebor@redhat.com>
11639 * attribs.c (decl_attributes): Reject conflicting attributes before
11640 calling attribute handlers.
11642 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
11644 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
11646 (gcc::dump_manager::get_dump_file_name): likewise.
11647 (dump_begin): Likewise.
11648 * dumpfile.h (dump_begin): Update prototype.
11649 (gcc::dump_manager::get_dump_file_name,
11650 gcc::dump_manager::get_dump_file_name): Update prototype.
11652 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
11654 * internal-fn.h (vectorizable_internal_fn_p): New function.
11655 * tree-vect-slp.c (compatible_calls_p): Likewise.
11656 (vect_build_slp_tree_1): Remove nops argument. Handle calls
11657 to internal functions.
11658 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
11660 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
11662 * fold-const.h (inverse_conditions_p): Declare.
11663 * fold-const.c (inverse_conditions_p): New function.
11664 * match.pd: Use inverse_conditions_p. Add folds of view_converts
11665 that test the inverse condition of a conditional internal function.
11666 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
11667 * internal-fn.c (internal_fn_mask_index): Handle conditional
11668 internal functions.
11669 (vectorized_internal_fn_supported_p): New function.
11670 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
11671 (any_pred_load_store): Replace with...
11672 (need_to_predicate): ...this new variable.
11673 (redundant_ssa_names): New variable.
11674 (ifcvt_can_use_mask_load_store): Move initial checks to...
11675 (ifcvt_can_predicate): ...this new function. Handle tree codes
11676 for which a conditional internal function exists.
11677 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
11678 instead of ifcvt_can_use_mask_load_store. Update after variable
11680 (predicate_load_or_store): New function, split out from
11681 predicate_mem_writes.
11682 (check_redundant_cond_expr): New function.
11683 (value_available_p): Likewise.
11684 (predicate_rhs_code): Likewise.
11685 (predicate_mem_writes): Rename to...
11686 (predicate_statements): ...this. Use predicate_load_or_store
11687 and predicate_rhs_code.
11688 (combine_blocks, tree_if_conversion): Update after above name changes.
11689 (ifcvt_local_dce): Handle redundant_ssa_names.
11690 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
11691 general conditional functions.
11692 * tree-vect-stmts.c (vectorizable_call): Likewise.
11694 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
11695 Alan Hayward <alan.hayward@arm.com>
11696 David Sherwood <david.sherwood@arm.com>
11698 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
11699 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
11700 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
11701 plus and minus and convert them into IFN_COND_FMA-based sequences.
11702 (convert_mult_to_fma): Handle conditional plus and minus.
11704 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
11706 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
11707 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
11708 (cond_fnms_optab): New optabs.
11709 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
11710 internal functions.
11711 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
11712 * internal-fn.h (get_conditional_internal_fn): Declare.
11713 (get_unconditional_internal_fn): Likewise.
11714 * internal-fn.c (cond_ternary_direct): New macro.
11715 (expand_cond_ternary_optab_fn): Likewise.
11716 (direct_cond_ternary_optab_supported_p): Likewise.
11717 (FOR_EACH_COND_FN_PAIR): Likewise.
11718 (get_conditional_internal_fn): New function.
11719 (get_unconditional_internal_fn): Likewise.
11720 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
11721 (gimple_match_op::gimple_match_op): Add a new overload for 5
11723 (gimple_match_op::set_op): Likewise.
11724 (gimple_resimplify5): Declare.
11725 * genmatch.c (decision_tree::gen): Generate simplifications for
11727 * gimple-match-head.c (gimple_simplify): Define an overload for
11728 5 operands. Handle calls with 5 arguments in the top-level overload.
11729 (convert_conditional_op): Handle conversions from unconditional
11730 internal functions to conditional ones.
11731 (gimple_resimplify5): New function.
11732 (build_call_internal): Pass a fifth operand.
11733 (maybe_push_res_to_seq): Likewise.
11734 (try_conditional_simplification): Try converting conditional
11735 internal functions to unconditional internal functions.
11736 Handle 3-operand unconditional forms.
11737 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
11738 Define ternary equivalents of the current rules for binary conditional
11739 internal functions.
11740 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
11741 ternary operations.
11742 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
11743 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
11744 (optab): Handle them.
11745 (SVE_COND_FP_TERNARY): New int iterator.
11746 (sve_fmla_op, sve_fmad_op): New int attributes.
11747 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
11748 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
11749 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
11751 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
11753 * target.def (preferred_else_value): New target hook.
11754 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
11755 * doc/tm.texi: Regenerate.
11756 * targhooks.h (default_preferred_else_value): Declare.
11757 * targhooks.c (default_preferred_else_value): New function.
11758 * internal-fn.h (conditional_internal_fn_code): Declare.
11759 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
11760 (get_conditional_internal_fn): Use it.
11761 (conditional_internal_fn_code): New function.
11762 * gimple-match.h (gimple_match_cond): New struct.
11763 (gimple_match_op): Add a cond member function.
11764 (gimple_match_op::gimple_match_op): Update all forms to take a
11766 * genmatch.c (expr::gen_transform): Use the same condition as res_op
11767 for the suboperation, but don't specify a particular else_value.
11768 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
11769 (visit_nary_op, visit_reference_op_load): Pass
11770 gimple_match_cond::UNCOND to the gimple_match_op constructor.
11771 * gimple-match-head.c: Include tree-eh.h
11772 (convert_conditional_op): New function.
11773 (maybe_resimplify_conditional_op): Likewise.
11774 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
11775 (gimple_resimplify2): Likewise.
11776 (gimple_resimplify3): Likewise.
11777 (gimple_resimplify4): Likewise.
11778 (maybe_push_res_to_seq): Return null for conditional operations.
11779 (try_conditional_simplification): New function.
11780 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
11782 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
11784 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
11786 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
11788 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
11790 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
11792 (hash_tree): Do not hash DECL_FCONTEXT
11793 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
11794 Do not stream DECL_FCONTEXT.
11795 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
11796 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
11798 2018-07-12 Richard Biener <rguenther@suse.de>
11801 * dwarf2out.c (gen_block_die): Only output blocks when they have
11802 at least one !DECL_IGNORED_P variable.
11804 2018-07-12 Richard Biener <rguenther@suse.de>
11807 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
11808 Remove -mieee-fp handling.
11810 2018-07-12 Richard Biener <rguenther@suse.de>
11812 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
11813 left-over from last patch.
11815 2018-07-12 Jakub Jelinek <jakub@redhat.com>
11817 PR tree-optimization/86492
11818 * gimple-ssa-store-merging.c
11819 (imm_store_chain_info::coalesce_immediate_stores): Call
11820 check_no_overlap even for the merge_overlapping case. Formatting fix.
11822 2018-07-12 Richard Biener <rguenther@suse.de>
11824 PR middle-end/86479
11825 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
11826 move possibly trapping operations into the conditional.
11828 2018-07-12 Richard Biener <rguenther@suse.de>
11830 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
11831 (vn_lookup_simplify_result): Remove recursion limit applied
11833 (vn_nary_build_or_lookup_1): Adjust.
11834 (try_to_simplify): Likewise.
11835 * gimple-match-head.c (gimple_resimplify1): Instead apply one
11837 (gimple_resimplify2): Likewise.
11838 (gimple_resimplify3): Likewise.
11839 (gimple_resimplify4): Likewise.
11841 2018-07-11 Jakub Jelinek <jakub@redhat.com>
11843 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
11844 Use __mmask64 type instead of __mmask8 for __M argument.
11845 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
11846 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
11848 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
11849 __mmask16 for __M argument.
11850 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
11851 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
11852 to __mmask16 instead of __mmask8.
11853 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
11854 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
11855 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
11856 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
11857 instead of __mmask16 for __U argument.
11858 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
11859 __mmask16 instead of __mmask8 for __U argument.
11860 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
11862 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
11864 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
11866 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
11868 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
11870 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
11872 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
11874 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
11876 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
11878 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
11880 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
11881 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
11882 return type as well as __M argument type and all casts from __mmask8
11884 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
11885 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
11886 return type as well as __M argument type and all casts from __mmask8
11888 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
11889 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
11890 return type as well as __M argument type and all casts from __mmask8
11892 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
11893 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
11894 return type as well as __M argument type and all casts from __mmask8
11896 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
11897 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
11900 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
11902 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
11903 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
11906 2018-07-11 Paul Koning <ni1d@arrl.net>
11908 * doc/md.texi (define_subst): Document how multiple occurrences of
11909 the same argument in the replacement pattern are handled.
11911 2018-07-11 Paul Koning <ni1d@arrl.net>
11913 * doc/extend.texi (Common Variable Attributes): Move "mode" into
11914 alphabetical order.
11915 (Common Type Attributes): Add "mode" attribute.
11917 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
11919 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
11920 stream DECL_ORIGINAL_TYPE.
11921 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
11922 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
11923 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
11924 Do not walk original type.
11925 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
11927 (write_ts_decl_non_common_tree_pointers): Do not stream
11929 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
11930 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
11932 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
11934 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
11935 thread twice from the same starting edge.
11937 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
11939 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
11940 * gimple.c (gimple_call_nonnull_result_p): ...here...
11941 (gimple_call_nonnull_arg): ...and here.
11942 * gimple.h (gimple_call_nonnull_result_p): New.
11943 (gimple_call_nonnull_arg): New.
11945 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
11947 * config/arm/arm-cpus.in: Move information from fpu field of each
11948 cpu definition to the isa field.
11949 * config/arm/parsecpu.awk (fpu): Delete match rule.
11950 (gen_comm_data): Don't add bits from the CPU's FPU entry.
11952 2018-07-11 Richard Biener <rguenther@suse.de>
11955 * dwarf2out.c (init_sections_and_labels): Use
11956 output_asm_line_debug_info consistently.
11957 (dwarf2out_early_finish): Likewise.
11958 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
11961 2018-07-11 Richard Biener <rguenther@suse.de>
11963 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
11964 Rework father_bb setting in a way to avoid propagating constants
11965 multiple times on a loop body.
11967 2018-07-10 Mark Wielaard <mark@klomp.org>
11970 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
11972 2018-07-10 Richard Biener <rguenther@suse.de>
11974 * hash-map.h (hash_map::iterator::operator*): Return
11975 references to key and value.
11977 2018-07-10 Jakub Jelinek <jakub@redhat.com>
11980 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
11981 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
11982 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
11983 (gimplify_omp_for): For composite loops, move outer
11984 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
11985 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
11986 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
11987 TREE_LIST for both the original class iterator and the "last" helper
11988 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
11989 loop, remember has_decl_expr from outer composite loops for the
11990 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
11992 2018-07-09 Martin Sebor <msebor@redhat.com>
11994 PR middle-end/77357
11995 PR middle-end/86428
11996 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
11997 accessing implicitly initialized array elements.
11998 * expr.c (string_constant): Handle string initializers of
11999 character arrays within aggregates.
12000 * gimple-fold.c (fold_array_ctor_reference): Add argument.
12001 Store element offset. As a special case, handle zero size.
12002 (fold_nonarray_ctor_reference): Same.
12003 (fold_ctor_reference): Add argument. Store subobject offset.
12004 * gimple-fold.h (fold_ctor_reference): Add argument.
12006 2018-07-09 Paul Koning <ni1d@arrl.net>
12008 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
12009 (pdp11_insn_cost): New function.
12010 (pdp11_md_asm_adjust): New function.
12011 (TARGET_INVALID_WITHIN_DOLOOP): Define.
12012 (pdp11_rtx_costs): Update to match machine better.
12013 (output_addr_const_pdp11): Correct format mismatch warnings.
12014 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
12015 * config/pdp11/pdp11.md: General change to add base_cost and/or
12016 length attributes for use by new pdp11_insn_cost function.
12017 (MIN_BRANCH): Correct definition.
12019 (doloop_end): Use standard pattern name for looping pattern.
12020 (doloop_end_nocc): New.
12021 (movsf): Add another constraint alternative.
12022 (zero_extendqihi2): Add constraint alternatives for not in place
12024 (zero_extendhisi2): Remove.
12025 (shift patterns): Add CC handling variants.
12029 (define_peephole2): New peephole to recognize mov that sets CC for
12032 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12034 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
12035 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
12038 2018-07-09 Jakub Jelinek <jakub@redhat.com>
12041 * real.c (real_nextafter): Return true if result is denormal.
12043 2018-07-09 Martin Liska <mliska@suse.cz>
12045 * common.opt: Add back wrongly removed attribute.
12047 2018-07-09 Richard Biener <rguenther@suse.de>
12050 * dwarf2out.c (gen_block_die): For an early generated DIE
12051 always output high/low PC attributes.
12053 2018-07-09 Tom de Vries <tdevries@suse.de>
12055 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
12056 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
12057 onto VAR_DECL with abstract origin.
12059 2018-07-07 Jim Wilson <jimw@sifive.com>
12061 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
12063 2018-07-07 Tom de Vries <tdevries@suse.de>
12065 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
12067 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
12069 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
12071 (combine_bound): Use wide-int overflow calculation instead of
12073 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
12075 * fold-const.c (int_const_binop_2): Same.
12076 (extract_muldiv_1): Same.
12077 (fold_div_compare): Same.
12078 (fold_abs_const): Same.
12080 * poly-int.h (add): Same.
12084 * predict.c (predict_iv_comparison): Same.
12085 * profile-count.c (slow_safe_scale_64bit): Same.
12086 * simplify-rtx.c (simplify_const_binary_operation): Same.
12087 * tree-chrec.c (tree_fold_binomial): Same.
12088 * tree-data-ref.c (split_constant_offset_1): Same.
12089 * tree-if-conv.c (idx_within_array_bound): Same.
12090 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
12091 * tree-ssa-phiopt.c (minmax_replacement): Same.
12092 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
12093 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
12094 * vr-values.c (vr_values::adjust_range_with_scev): Same.
12095 * wide-int.cc (wi::add_large): Same.
12096 (wi::mul_internal): Same.
12097 (wi::sub_large): Same.
12098 (wi::divmod_internal): Same.
12099 * wide-int.h: Change overflow type to overflow_type for neg, add,
12100 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
12101 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
12102 mul_internal, divmod_internal.
12103 (overflow_type): New enum.
12104 (accumulate_overflow): New.
12106 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
12108 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
12109 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
12111 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
12113 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
12114 argument is checked for zero before entering loop, avoid checking again.
12116 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
12118 * gimplify.h (generic_expr_could_trap_p): Set as global function.
12119 * gimplify.h (generic_expr_could_trap_p): Likwise.
12120 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
12122 2018-07-06 Jakub Jelinek <jakub@redhat.com>
12124 PR tree-optimization/86401
12125 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
12126 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
12127 (fold_bit_and_mask): ... here. New helper function for match.pd.
12128 * fold-const.h (fold_bit_and_mask): Declare.
12129 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
12131 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
12134 * target.def (translate_mode_attribute): New hook.
12135 * targhooks.h (default_translate_mode_attribute): Declare.
12136 * targhooks.c (default_translate_mode_attribute): New function.
12137 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
12138 * doc/tm.texi: Regenerate.
12139 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
12140 (rs6000_translate_mode_attribute): New function.
12142 2018-07-06 Paul Koning <ni1d@arrl.net>
12144 * doc/md.texi (define_split): Document DONE and FAIL.
12145 (define_peephole2): Ditto.
12147 2018-07-05 Jeff Law <law@redhat.com>
12149 PR tree-optimization/86010
12150 * tree-ssa-dse.c (compute_trims): More aggressively trim at
12151 both the head and tail of mem* and str* calls.
12153 2018-07-05 Jim Wilson <jimw@sifive.com>
12155 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
12157 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
12159 * config/aarch64/aarch64-simd.md: correct flags text for
12160 MIN_EXPR replacement.
12162 2018-07-05 James Clarke <jrtc27@jrtc27.com>
12164 * configure: Regenerated.
12166 2018-07-05 Carl Love <cel@us.ibm.com>
12168 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
12169 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
12170 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
12171 VSX_BUILTIN_DOUBLEL_V4SF.
12173 2018-07-05 Martin Sebor <msebor@redhat.com>
12176 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
12177 than its domain to compute its the upper bound of a char array.
12179 2018-07-05 Nathan Sidwell <nathan@acm.org>
12181 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
12182 * doc/cpp.texi: Update comment.
12183 * doc/tm.texi: Rebuilt.
12184 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
12185 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
12186 * doc/extend.texi (Backwards Compatibility): Clarify it is system
12187 headers affected by extern "C".
12188 * system.h: Poison NO_IMPLICIT_EXTERN_C.
12189 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
12190 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
12191 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
12192 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
12193 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
12194 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
12195 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
12196 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
12197 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
12198 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
12199 config/sparc/sp64-elf.h, config/spu/spu.h,
12200 config/stormy16/stormy16.h, config/v850/v850.h,
12201 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
12202 define NO_IMPLICIT_EXTERN_C.
12203 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
12205 2018-07-05 Tamar Christina <tamar.christina@arm.com>
12208 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
12209 instead of GET_MODE_SIZE when comparing Units.
12211 2018-07-05 Tamar Christina <tamar.christina@arm.com>
12214 * rtlanal.c (set_noop_p): Constrain on mode change,
12215 include hard-reg-set.h
12217 2018-07-05 Tamar Christina <tamar.christina@arm.com>
12219 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
12221 2018-07-05 Jakub Jelinek <jakub@redhat.com>
12224 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
12227 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
12229 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
12231 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
12234 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
12236 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
12238 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
12240 PR middle-end/86380
12241 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
12243 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
12245 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
12246 neg_*_op* variables.
12248 2018-07-04 Martin Liska <mliska@suse.cz>
12250 * tree-switch-conversion.c: Define
12251 max_ratio_for_speed and max_ratio_for_size constants.
12253 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
12254 Martin Liska <mliska@suse.cz>
12256 PR middle-end/66240
12259 * common.opt: Rename align options with 'str_' prefix.
12260 * common/config/i386/i386-common.c (set_malign_value): New
12262 (ix86_handle_option): Use it to set -falign-* options/
12263 * config/aarch64/aarch64-protos.h (struct tune_params): Change
12264 type from int to string.
12265 * config/aarch64/aarch64.c: Update default values from int
12267 * config/alpha/alpha.c (alpha_override_options_after_change):
12269 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
12270 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12271 max skip conditionally.
12272 * config/i386/freebsd.h (SUBALIGN_LOG): New.
12273 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12274 max skip conditionally.
12275 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12276 max skip conditionally.
12277 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
12278 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12279 max skip conditionally.
12280 * config/i386/i386.c (struct ptt): Change type from int to
12282 (ix86_default_align): Set default values.
12283 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
12284 max skip conditionally.
12285 * config/i386/iamcu.h (SUBALIGN_LOG): New.
12286 (ASM_OUTPUT_MAX_SKIP_ALIGN):
12287 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
12288 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12289 max skip conditionally.
12290 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
12291 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
12292 * config/i386/x86-64.h (SUBALIGN_LOG): New.
12293 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12294 max skip conditionally.
12295 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12296 * config/ia64/ia64.c (ia64_option_override): Set default values
12297 for alignment options.
12298 * config/m68k/m68k.c: Handle new str_align_* options.
12299 * config/mips/mips.c (mips_set_compression_mode): Change
12301 (mips_option_override): Set default values for options.
12302 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
12304 * config/rs6000/rs6000.c (rs6000_option_override_internal):
12306 * config/rx/rx.c (rx_option_override): Likewise.
12307 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
12308 (LABEL_ALIGN): Use align_labels_log.
12309 (LOOP_ALIGN): Use align_loops_align.
12310 * config/s390/s390.c (s390_asm_output_function_label): Use new
12312 * config/sh/sh.c (sh_override_options_after_change):
12313 Change type of constants.
12314 * config/spu/spu.c (spu_sched_init): Likewise.
12315 * config/sparc/sparc.c (sparc_option_override): Set default
12316 values for options.
12317 * config/visium/visium.c (visium_option_override): Likewise.
12318 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
12319 emit p2align format with last argument if it's not needed.
12320 * doc/invoke.texi: Document extended format of -falign-*.
12321 * final.c: Use align_labels alignment.
12322 * flags.h (struct target_flag_state): Change type to use
12324 (struct align_flags_tuple): New.
12325 (struct align_flags): Likewise.
12326 (align_loops_log): Redefine macro to use new types.
12327 (align_loops_max_skip): Redefine macro to use new types.
12328 (align_jumps_log): Redefine macro to use new types.
12329 (align_jumps_max_skip): Redefine macro to use new types.
12330 (align_labels_log): Redefine macro to use new types.
12331 (align_labels_max_skip): Redefine macro to use new types.
12332 (align_functions_log): Redefine macro to use new types.
12333 (align_loops): Redefine macro to use new types.
12334 (align_jumps): Redefine macro to use new types.
12335 (align_labels): Redefine macro to use new types.
12336 (align_functions): Redefine macro to use new types.
12337 (align_functions_max_skip): Redefine macro to use new types.
12338 (align_loops_value): New macro.
12339 (align_jumps_value): New macro.
12340 (align_labels_value): New macro.
12341 (align_functions_value): New macro.
12342 * function.c (invoke_set_current_function_hook): Propagate
12343 alignment values from flags to global variables default in
12345 * ipa-icf.c (sem_function::equals_wpa): Use
12346 cl_optimization_option_eq instead of memcmp.
12347 * lto-streamer.h (cl_optimization_stream_out): Support streaming
12349 (cl_optimization_stream_in): Likewise.
12350 * optc-save-gen.awk: Support strings in cl_optimization.
12351 * opth-gen.awk: Likewise.
12352 * opts.c (finish_options): Remove error checking of invalid
12354 (MAX_CODE_ALIGN): Remove.
12355 (MAX_CODE_ALIGN_VALUE): Likewise.
12356 (parse_and_check_align_values): New function.
12357 (check_alignment_argument): Likewise.
12358 (common_handle_option): Use check_alignment_argument.
12359 * opts.h (parse_and_check_align_values): Declare.
12360 * toplev.c (init_alignments): Remove.
12361 (read_log_maxskip): New.
12362 (parse_N_M): Likewise.
12363 (parse_alignment_opts): Likewise.
12364 (backend_init_target): Remove usage of init_alignments.
12365 * toplev.h (parse_alignment_opts): Declare.
12366 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
12368 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
12369 * tree.c (cl_option_hasher::equal): New.
12370 * varasm.c: Use new global macros.
12372 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
12374 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
12375 Use a simpler align directive also if MAXSKIP = ALIGN-1.
12376 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12377 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12378 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12379 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12380 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
12381 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
12382 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
12383 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12384 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12385 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12386 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12388 2018-07-04 Martin Liska <mliska@suse.cz>
12389 Jonathan Wakely <jwakely@redhat.com>
12391 * coverage.c: Use correct type.
12392 * doc/invoke.texi: Language correction.
12394 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
12397 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
12398 ENDBRANCH for non-tail call which may return via indirect branch.
12399 * doc/extend.texi: Document indirect_return attribute.
12401 2018-07-03 Martin Sebor <msebor@redhat.com>
12403 PR tree-optimization/86274
12404 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
12406 (format_floating): Correct handling of infinities and NaNs.
12408 2018-07-03 Martin Sebor <msebor@redhat.com>
12410 * print-tree.c (print_real_cst): New function.
12411 (print_node_brief): Call it.
12412 (print_node): Ditto.
12414 2018-07-03 Jeff Law <law@redhat.com>
12416 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
12417 into a single pattern.
12419 * config/h8300/h8300.md (ors code_iterator): New.
12420 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
12421 a single pattern and single splitter.
12422 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
12423 (iorqi3_1, xorqi3_1): Likewise.
12424 (iorqi3, xorqi3 expanders): Similarly.
12426 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
12427 (movmd_internal) into a single pattern using the P mode iterator.
12428 (movmd splitters): Similarly.
12429 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
12430 (movsd splitters): Similarly.
12432 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
12433 ADDB, ADDW and ADDL into a single ADD attribute which selects the
12434 right table based on the size of the operand.
12435 * config/h8300/h8300.md (length_table): Corresponding changes. All
12436 references to "addb", "addw" and "addl" changed to "add".
12437 (btst patterns): Merge two variants into a single pattern.
12438 (tstqi, tsthi): Likewise.
12439 (addhi3_incdec, addsi3_incdec): Likewise.
12440 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
12441 (mulhi3, mulsi3): Likewise.
12442 (udivhi3, udivsi3): Likewise.
12443 (divhi3, divsi3): Likewise.
12444 (andorqi3, andorhi3, andorsi3): Likewise.
12446 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
12449 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
12450 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
12452 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
12454 PR tree-optimization/85694
12455 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
12456 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
12458 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
12459 (<u>avg<mode>3_ceil): New patterns.
12461 2018-07-03 David Malcolm <dmalcolm@redhat.com>
12463 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
12464 scan-tree-dump directive.
12465 * gcc.dg/vect/slp-perm-2.c: Likewise.
12466 * gcc.dg/vect/slp-perm-3.c: Likewise.
12467 * gcc.dg/vect/slp-perm-5.c: Likewise.
12468 * gcc.dg/vect/slp-perm-6.c: Likewise.
12469 * gcc.dg/vect/slp-perm-7.c: Likewise.
12470 * gcc.dg/vect/slp-perm-8.c: Likewise.
12472 2018-07-03 Marek Polacek <polacek@redhat.com>
12474 PR middle-end/86202
12475 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
12477 2018-07-03 Richard Biener <rguenther@suse.de>
12480 * tree-ssa-structalias.c (find_func_clobbers): Properly
12481 handle indirect calls.
12483 2018-07-03 Jeff Law <law@redhat.com>
12485 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
12486 (shifts): New code iterator.
12487 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
12488 expander. Fix HImode handling on H8/SX.
12489 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
12490 (subqi3, subhi3, subsi3 expanders): Likewise.
12491 (andqi3, andhi3, andsi3 expanders): Likewise.
12492 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
12493 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
12494 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
12495 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
12496 (zero_extendqihi2, zero_extendqisi2): Likewise.
12497 (extendqihi2, extendqisi2): Likewise.
12498 (rotlqi3, rotlhi3, rotlsi3): Likewise.
12499 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
12500 (rotlqi3_1, rotlhi3_1): Likewise.
12501 (logicalhi3_sn, logicalsi3_sn): Likewise.
12502 (logicalhi3, logicalsi3): Likewise.
12504 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
12506 * tree-vect-patterns.c (vect_recog_rotate_pattern)
12507 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
12508 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
12509 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
12510 type to append_pattern_def_seq instead of creating a stmt_vec_info
12512 (build_mask_conversion): Likewise. Remove vinfo argument.
12513 (vect_add_conversion_to_patterm): Likewise, renaming to...
12514 (vect_add_conversion_to_pattern): ...this.
12515 (vect_recog_mask_conversion_pattern): Update call to
12516 build_mask_conversion. Pass the vector type to
12517 append_pattern_def_seq here too.
12518 (vect_recog_gather_scatter_pattern): Update call to
12519 vect_add_conversion_to_pattern.
12521 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
12523 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
12524 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
12525 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
12526 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
12527 STMT_VINFO_PATTERN_DEF_SEQ to null here.
12528 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
12529 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
12530 append_pattern_def_seq instead of new_pattern_def_seq.
12531 (vect_recog_divmod_pattern): Do both of the above.
12532 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
12535 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
12537 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
12538 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
12539 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
12540 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
12541 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
12542 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
12543 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
12544 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
12545 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
12546 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
12547 parameter with a single stmt_vec_info.
12548 (vect_recog_func_ptr): Likewise.
12549 (vect_recog_gather_scatter_pattern): Likewise, folding in...
12550 (vect_try_gather_scatter_pattern): ...this.
12551 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
12552 the stmt_vec_info of the statement to be matched. Don't clear
12553 STMT_VINFO_RELATED_STMT.
12554 (vect_pattern_recog): Update call accordingly.
12556 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
12558 PR tree-optimization/85694
12559 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
12560 (uavgM3_ceil): Document new optabs.
12561 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
12562 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
12564 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
12565 (savg_ceil_optab): New optabs.
12566 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
12567 (vect_vect_recog_func_ptrs): Add it.
12568 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
12569 constant directly from the associated lhs.
12571 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
12573 * tree-vect-patterns.c (vect_split_statement): New function.
12574 (vect_convert_input): Use it to try to split an existing cast.
12576 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
12578 * poly-int.h (print_hex): New function.
12579 * dumpfile.h (dump_dec, dump_hex): Declare.
12580 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
12581 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
12582 min_input_precision, operation_precision and operation_sign.
12583 * tree-vect-patterns.c (vect_get_range_info): New function.
12584 (vect_same_loop_or_bb_p, vect_single_imm_use)
12585 (vect_operation_fits_smaller_type): Delete.
12586 (vect_look_through_possible_promotion): Add an optional
12587 single_use_p parameter.
12588 (vect_recog_over_widening_pattern): Rewrite to use new
12589 stmt_vec_info infomration. Handle one operation at a time.
12590 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
12591 (vect_truncatable_operation_p, vect_set_operation_type)
12592 (vect_set_min_input_precision): New functions.
12593 (vect_determine_min_output_precision_1): Likewise.
12594 (vect_determine_min_output_precision): Likewise.
12595 (vect_determine_precisions_from_range): Likewise.
12596 (vect_determine_precisions_from_users): Likewise.
12597 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
12598 (vect_vect_recog_func_ptrs): Put over_widening first.
12600 (vect_pattern_recog): Call vect_determine_precisions.
12602 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
12604 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
12605 statements that have been replaced by further pattern statements.
12606 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
12608 2018-07-03 Richard Biener <rguenther@suse.de>
12610 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
12611 always set *dt. Dump vectype in vectype overload.
12612 * dumpfile.h (dump_gimple_expr): New function.
12613 (dump_gimple_expr_loc): Likewise.
12614 * dumpfile.c (dump_gimple_expr): New function.
12615 (dump_gimple_expr_loc): Likewise.
12617 2018-07-02 Jeff Law <law@redhat.com>
12619 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
12620 the H8/300, H8/300H and H8/S variants into a single pattern.
12621 (movhi_h8300, movqi_h8300hs): Similarly.
12622 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
12623 (QHI mode iterator): New.
12625 * config/h8300/h8300.md: Remove trailing whitespace.
12627 2018-07-02 Jim Wilson <jimw@sifive.com>
12629 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
12630 instead of emit_insn for interrupt returns.
12631 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
12632 (riscv_sret, riscv_uret): Likewise.
12634 2018-07-02 David Malcolm <dmalcolm@redhat.com>
12636 * pretty-print.c (selftest::test_pp_format): Move save and restore
12637 of quotes to class auto_fix_quotes, and add an instance.
12638 * selftest.c: Include "intl.h".
12639 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
12640 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
12641 * selftest.h (selftest::auto_fix_quotes): New class.
12643 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
12645 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
12646 (aarch64_sve_prepare_conditional_op): Remove.
12647 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
12648 Allow aarch64_simd_reg_or_zero as select operand; remove
12649 the aarch64_sve_prepare_conditional_op call.
12650 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
12651 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
12652 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
12653 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
12654 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
12655 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
12656 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
12657 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
12658 and a splitters to match all of the *_any patterns.
12659 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
12661 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
12662 (SVE_COND_FP_BINARY_REV): Remove.
12663 (sve_int_op_rev, sve_fp_op_rev): New.
12664 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
12665 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
12666 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
12667 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
12668 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
12669 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
12670 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
12671 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
12672 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
12674 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
12675 Remove match_dup 1 from the inner unspec.
12676 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
12678 * config/aarch64/aarch64.md (movprfx): New attr.
12679 (length): Default movprfx to 8.
12680 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
12681 (*madd<SVE_I>, *msub<SVE_I): Likewise.
12682 (*<su>mul<SVE_I>3_highpart): Likewise.
12683 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
12684 (*v<ASHIFT><SVE_I>3): Likewise.
12685 (*<su><MAXMIN><SVE_I>3): Likewise.
12686 (*<su><MAXMIN><SVE_F>3): Likewise.
12687 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
12688 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
12689 (*div<SVE_F>4): Likewise.
12691 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
12693 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
12696 2018-07-02 Richard Biener <rguenther@suse.de>
12698 PR tree-optimization/86363
12699 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
12700 memset argument refers to a non-variable address.
12702 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
12704 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
12705 {PLUS,MINUS}_EXPR code to...
12706 (adjust_symbolic_bound): ...here,
12707 (combine_bound): ...here,
12708 (set_value_range_with_overflow): ...and here.
12710 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
12712 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
12714 (extract_range_from_abs_expr): ...here.
12716 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
12718 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
12719 -fno-omit-frame-pointer when not optimizing.
12721 2018-07-02 Martin Liska <mliska@suse.cz>
12724 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
12725 (propagate_nothrow): Likewise.
12727 2018-07-02 Martin Liska <mliska@suse.cz>
12730 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
12732 2018-07-02 David Malcolm <dmalcolm@redhat.com>
12734 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
12735 function in r262149, changing "loc" param from source_location to
12736 const dump_location_t &.
12737 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
12738 declaration, as above.
12740 2018-07-01 Paul Koning <ni1d@arrl.net>
12742 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
12743 -munit-asm, -mgnu-asm, -mdec-asm.
12744 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
12745 (pdp11_output_labelref): New.
12746 (pdp11_output_def): New.
12747 (pdp11_output_addr_vec_elt): New.
12748 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
12749 %# and %@ format codes.
12750 (pdp11_option_override): New.
12751 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
12752 (pdp11_output_ident): New.
12753 (pdp11_asm_named_section): New.
12754 (pdp11_asm_init_sections): New.
12755 (pdp11_file_start): New.
12756 (pdp11_file_end): New.
12757 (output_ascii): Use .ascii/.asciz for -mdec-asm.
12758 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
12759 %o, like %c but octal.
12760 (pdp11_option_override): New.
12761 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
12763 (DATA_SECTION_ASM_OP): Ditto.
12764 (READONLY_DATA_SECTION_ASM_OP): New.
12765 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
12766 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
12767 (ASM_OUTPUT_LABELREF): Ditto.
12768 (ASM_OUTPUT_DEF): Ditto.
12769 (ASM_OUTPUT_EXTERNAL): New.
12770 (ASM_OUTPUT_SOURCE_FILENAME): New.
12771 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
12772 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
12773 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
12774 %# and %@ format codes.
12775 * config/pdp11/pdp11.opt (mgnu-asm): New.
12776 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
12777 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
12778 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
12780 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
12782 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
12783 dereferencing path[] beyond its length.
12785 (debug_all_paths): New.
12786 (rewire_first_differing_edge): New.
12787 (adjust_paths_after_duplication): New.
12788 (duplicate_thread_path): Call adjust_paths_after_duplication.
12790 (thread_through_all_blocks): Add new argument to
12791 duplicate_thread_path.
12793 2018-06-30 Jim Wilson <jimw@sifive.com>
12795 * config/riscv/predicates.md (p2m1_shift_operand): New.
12796 (high_mask_shift_operand): New.
12797 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
12798 pattern using p2m1_shift_operand.
12799 (lshsi3_zero_extend_3+2): New combiner pattern using
12800 high_mask_shift_operand.
12802 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
12804 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
12806 (vect_recog_rotate_pattern): ...here.
12807 (vect_convert_input): Try to insert casts of invariants in the
12809 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
12810 preheader to be empty.
12812 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
12814 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
12815 vector type. If given, install it in the new statement's
12816 STMT_VINFO_VECTYPE.
12817 (vect_element_precision): New function.
12818 (vect_unpromoted_value): New struct.
12819 (vect_unpromoted_value::vect_unpromoted_value): New function.
12820 (vect_unpromoted_value::set_op): Likewise.
12821 (vect_look_through_possible_promotion): Likewise.
12822 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
12823 (vect_widened_op_tree, vect_convert_input): Likewise.
12824 (vect_convert_inputs, vect_convert_output): Likewise.
12825 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
12826 to handle the optional cast of the multiplication result and
12827 vect_widened_op_tree to detect the widened multiplication itself.
12828 Do not require the input and output of promotion casts to have
12829 the same sign, but base the signedness of the operation on the
12830 input rather than the result. If the pattern includes two
12831 promotions, check that those promotions have the same sign.
12832 Do not restrict the MULT_EXPR handling to a double-width result;
12833 handle quadruple-width results and wider. Use vect_convert_inputs
12834 to convert the inputs to the common type.
12835 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
12836 to handle the optional cast of the ABS result. Also allow a sign
12837 change or a sign extension between the ABS and MINUS.
12838 Use vect_widened_op_tree to detect the widened subtraction and use
12839 vect_convert_inputs to convert the inputs to the common type.
12840 (vect_handle_widen_op_by_const): Delete.
12841 (vect_recog_widen_op_pattern): New function.
12842 (vect_recog_widen_mult_pattern): Use it.
12843 (vect_recog_widen_shift_pattern): Likewise.
12844 (vect_recog_widen_sum_pattern): Use
12845 vect_look_through_possible_promotion to handle the promoted
12848 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
12850 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
12851 the containing gimple_seq *.
12852 * gimple-iterator.h (gsi_for_stmt): Declare it.
12853 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
12854 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
12855 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
12856 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
12857 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
12859 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
12861 (vect_mark_pattern_stmts): ...here. Handle cases in which the
12862 statement being replaced is part of an existing pattern
12863 definition sequence, inserting the new pattern statements before
12865 (vect_pattern_recog_1): Don't return a bool. If the statement
12866 is already part of a pattern, instead apply pattern matching
12867 to the pattern definition statements. Don't clear the
12868 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
12869 (vect_pattern_recog): Don't break after the first match;
12870 continue processing the pattern definition statements instead.
12871 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
12873 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
12875 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12876 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
12877 (vect_recog_widen_sum_pattern): Use it.
12879 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
12881 * tree-vect-loop.c (vectorizable_reduction): Assert that the
12882 phi is not a pattern statement and has not been replaced by
12883 a pattern statement.
12884 * tree-vect-patterns.c (type_conversion_p): Don't check
12885 STMT_VINFO_IN_PATTERN_P.
12886 (vect_recog_vector_vector_shift_pattern): Likewise.
12887 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
12888 the pattern statement rather than the original statement; check
12889 directly for a WIDEN_MULT_EXPR here.
12890 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
12891 vect_is_simple_use to return the pattern statement rather
12892 than the original statement; use is_pattern_stmt_p to check
12893 for such a pattern statement.
12894 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
12895 to return the pattern statement rather than the original statement;
12896 don't do the same transformation here.
12897 (vect_is_simple_use): If the defining statement has been replaced
12898 by a pattern statement, return the pattern statement instead.
12899 Remove the corresponding (local) transformation from the vectype
12902 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
12904 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
12905 end and default to null.
12906 * tree-vect-loop.c (vect_create_epilog_for_reduction)
12907 (vectorizable_reduction): Update calls accordingly, dropping the
12908 gimple ** argument if the passed-back statement isn't needed.
12909 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
12910 (vect_recog_rotate_pattern): Likewise.
12911 (vect_recog_mask_conversion_pattern): Likewise.
12912 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
12913 (vect_mask_constant_operand_p): Likewise.
12914 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
12915 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
12916 (get_group_load_store_type, get_load_store_type): Likewise.
12917 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
12918 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
12919 (vectorizable_conversion, vectorizable_assignment): Likewise.
12920 (vectorizable_shift, vectorizable_operation): Likewise.
12921 (vectorizable_store, vect_is_simple_cond): Likewise.
12922 (vectorizable_condition, vectorizable_comparison): Likewise.
12923 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
12924 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
12925 and move it to the end. Cope with null def_stmt_outs.
12927 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
12929 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
12931 2018-06-29 Jeff Law <law@redhat.com>
12933 * config/v850/v850.c (v850_legitimate_address_p): Handle large
12934 displacements for TARGET_V850E2V3 and newer.
12935 (TARGET_LRA_P): Remove. Defaults to LRA now.
12936 * config/v850/v850.md (sign23byte_load): Remove.
12937 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
12938 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
12940 2018-06-29 Martin Liska <mliska@suse.cz>
12943 * coverage.c (coverage_init): Mangle full path name.
12944 * doc/invoke.texi: Document the change.
12945 * gcov-io.c (mangle_path): New.
12946 * gcov-io.h (mangle_path): Likewise.
12947 * gcov.c (mangle_name): Use mangle_path for path mangling.
12949 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12951 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
12952 if starting source register is not even.
12954 2018-06-29 Martin Liska <mliska@suse.cz>
12956 PR tree-optimization/86263
12957 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
12958 Make edge redirection.
12960 2018-06-29 David Malcolm <dmalcolm@redhat.com>
12962 * dumpfile.c (dump_loc): Add indentation based on scope depth.
12963 (dump_scope_depth): New variable.
12964 (get_dump_scope_depth): New function.
12965 (dump_begin_scope): New function.
12966 (dump_end_scope): New function.
12967 * dumpfile.h (get_dump_scope_depth): New declaration.
12968 (dump_begin_scope): New declaration.
12969 (dump_end_scope): New declaration.
12970 (class auto_dump_scope): New class.
12971 (AUTO_DUMP_SCOPE): New macro.
12972 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
12975 2018-06-29 Richard Biener <rguenther@suse.de>
12977 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
12978 compute_all_dependences succeeds.
12979 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
12980 exceed --param loop-max-datarefs-for-datadeps.
12982 2018-06-29 Jakub Jelinek <jakub@redhat.com>
12984 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
12986 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
12989 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
12990 alternative 0 in preferred_for_speed attribute.
12992 2018-06-28 Paul Koning <ni1d@arrl.net>
12994 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
12995 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
12996 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
12997 * config/pdp11/pdp11.md: Correct "length" attribute calculation
12998 for shift insn patterns.
13000 2018-06-28 David Malcolm <dmalcolm@redhat.com>
13002 * cgraph.c (cgraph_node::get_body): Replace assignments to
13003 "dump_file" with calls to set_dump_file.
13004 * dumpfile.c (alt_dump_file): Make static, and group with...
13005 (alt_flags): ...this definition.
13006 (dumps_are_enabled): New variable.
13007 (refresh_dumps_are_enabled): New function.
13008 (set_dump_file): New function.
13009 (set_alt_dump_file): New function.
13010 (gcc::dump_manager::dump_start): Replace assignments to
13011 "dump_file" and "alt_dump_file" with calls to set_dump_file and
13013 (gcc::dump_manager::dump_finish): Likewise.
13014 * dumpfile.h (alt_dump_file): Delete decl.
13015 (dumps_are_enabled): New variable decl.
13016 (set_dump_file): New function decl.
13017 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
13019 * tree-nested.c (lower_nested_functions): Replace assignments to
13020 "dump_file" with calls to set_dump_file.
13022 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
13024 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
13025 goto_locus of each outgoing edge of each basic block.
13027 2018-06-28 Richard Biener <rguenther@suse.de>
13029 * dwarf2out.c (decl_scope_table): Remove.
13030 (push_decl_scope): Likewise.
13031 (pop_decl_scope): Likewise.
13032 (gen_type_die_for_member): Do not call push/pop_decl_scope.
13033 (gen_struct_or_union_type_die): Likewise.
13034 (gen_tagged_type_die): Likewise.
13035 (dwarf2out_init): Do not initialize decl_scope_table.
13036 (dwarf2out_c_finalize): Do not free it.
13038 2018-06-28 Richard Biener <rguenther@suse.de>
13040 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
13041 deciding whether to not re-use a DIE.
13043 2018-06-28 Richard Biener <rguenther@suse.de>
13045 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
13046 DW_AT_abstract_origin attribute.
13048 2018-06-28 Martin Liska <mliska@suse.cz>
13050 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
13051 Use newly introduced constants.
13052 * tree-switch-conversion.h (struct jump_table_cluster):
13053 Define max_ratio_for_size and max_ratio_for_speed.
13055 2018-06-28 Martin Liska <mliska@suse.cz>
13057 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
13058 Add new checking assert to catch invalid state.
13059 (jump_table_cluster::can_be_handled): Handle single case
13061 (jump_table_cluster::is_beneficial): Bail out for such case.
13062 (bit_test_cluster::find_bit_tests):
13063 Add new checking assert to catch invalid state.
13064 (bit_test_cluster::can_be_handled): Handle single case
13066 (bit_test_cluster::is_beneficial): Bail out for such case.
13067 (switch_decision_tree::analyze_switch_statement):
13070 2018-06-28 Martin Liska <mliska@suse.cz>
13072 * common.opt: Introduce -completion option.
13073 * gcc.c (driver_handle_option): Handle it.
13074 (driver::main): Print completions if completion
13076 * opt-suggestions.c (option_proposer::get_completions):
13078 (option_proposer::suggest_completion): Likewise.
13079 (option_proposer::find_param_completions): Likewise.
13080 (verify_autocompletions): Likewise.
13081 (test_completion_valid_options): Likewise.
13082 (test_completion_valid_params): Likewise.
13083 (in_completion_p): Likewise.
13084 (empty_completion_p): Likewise.
13085 (test_completion_partial_match): Likewise.
13086 (test_completion_garbage): Likewise.
13087 (opt_proposer_c_tests): Likewise.
13088 * opt-suggestions.h: Declare new functions.
13089 * opts.c (common_handle_option): Handle OPT__completion_.
13090 * selftest-run-tests.c (selftest::run_tests): Add
13091 opt_proposer_c_tests.
13092 * selftest.c (assert_str_startswith): New.
13093 * selftest.h (assert_str_startswith): Likewise.
13094 (opt_proposer_c_tests): New.
13095 (ASSERT_STR_STARTSWITH): Likewise.
13097 2018-06-28 Martin Liska <mliska@suse.cz>
13099 * Makefile.in: Add opt-suggestions.o.
13100 * gcc-main.c: Include opt-suggestions.h.
13101 * gcc.c (driver::driver): Likewise.
13102 (driver::~driver): Remove m_option_suggestions.
13103 (driver::build_option_suggestions): Moved to option_proposer.
13104 (driver::suggest_option): Likewise.
13105 (driver::handle_unrecognized_options): Use option_proposer.
13106 * gcc.h (class driver): Add new memver m_option_proposer.
13107 * opt-suggestions.c: New file.
13108 * opt-suggestions.h: New file.
13110 2018-06-28 Martin Liska <mliska@suse.cz>
13112 * vec.h (class auto_string_vec): New (moved from auto_argvec).
13113 (auto_string_vec::~auto_string_vec): Likewise.
13115 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
13117 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
13118 prevent_decl_creation_for_types fields up and add reset_location field.
13119 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
13120 statement if id->reset_location is true.
13121 (copy_edges_for_bb): Do not set goto_locus on the new edges if
13122 id->reset_location is true.
13123 (copy_phis_for_bb): Force input_location on the arguments if
13124 id->reset_location is true.
13125 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
13126 is set on the function to be inlined.
13128 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
13130 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
13132 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
13134 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
13135 registers for Pmode.
13136 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
13137 hard registers for the clobbered pseudo.
13139 2018-06-27 Paul Koning <ni1d@arrl.net>
13141 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
13142 mutually exclusive options.
13143 * config/pdp11/constraints.md (h): New constraint.
13144 (O): Update definition to match shift code generation.
13145 (D): New constraint.
13146 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
13148 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
13150 (output_jump): Change arguments.
13151 (pdp11_fixed_cc_regs): New function.
13152 (pdp11_cc_mode): Ditto.
13153 (pdp11_expand_shift): Ditto.
13154 (pdp11_assemble_shift): Ditto.
13155 (pdp11_small_shift): Ditto.
13156 (pdp11_branch_cost): Remove.
13157 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
13159 (pdp11_register_move_cost): Update for CC registers.
13160 (pdp11_rtx_costs): Add case for LSHIFTRT.
13161 (pdp11_output_jump): Add CCNZ mode conditional branches.
13162 (notice_update_cc_on_set): Remove.
13163 (pdp11_cc_mode): New function.
13164 (simple_memory_operand): Correct pre/post decrement case.
13165 (no_side_effect_operand): New function.
13166 (pdp11_regno_reg_class): Add CC_REGS class.
13167 (pdp11_fixed_cc_regs): New function.
13168 (pdp11_small_shift): New function.
13169 (pdp11_expand_shift): New function to expand shift insns.
13170 (pdp11_assemble_shift): New function to output shifts.
13171 (pdp11_branch_cost): Remove.
13172 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
13173 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
13174 (WCHAR_TYPE): Ditto.
13175 (PTRDIFF_TYPE): Ditto.
13176 (ADJUST_INSN_LENGTH): New macro.
13177 (FIXED_REGISTERS): Add CC registers.
13178 (CALL_USED_REGISTERS): Ditto.
13179 (reg_class): Ditto.
13180 (REG_CLASS_NAMES): Ditto.
13181 (REG_CLASS_CONTENTS): Ditto.
13182 (SELECT_CC_MODE): Use new function.
13183 (TARGET_FLAGS_REGNUM): New macro.
13184 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
13185 (cc0_reg_rtx): Remove.
13186 (CC_STATUS_MDEP): Remove.
13187 (CC_STATUS_MDEFP_INIT): Remove.
13188 (CC_IN_FPU): Remove.
13189 (NOTICE_UPDATE_CC): Remove.
13190 (REGISTER_NAMES): Add CC registers.
13191 (BRANCH_COST): Change to constant 1.
13192 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
13194 * config/pdp11/pdp11.opt (mbcopy): Remove.
13195 (mbcopy-builtin): Remove.
13196 (mbranch-cheap): Remove.
13197 (mbranch-expensive): Remove.
13198 * config/pdp11/predicates.md (expand_shift_operand): Update to
13199 match shift code generation.
13200 (ccnz_operator): New predicate.
13201 * doc/invoke.texi (PDP-11 Options): Remove deleted options
13202 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
13203 Remove non-existent option -mabshi, -mno-abshi. Document mutually
13205 * doc/md.texi (PDP-11): Document new D and h constraints. Update
13206 description of O constraint.
13208 2018-06-27 Jeff Law <law@redhat.com>
13209 Austin Law <austinklaw@gmail.com>
13211 * config/v850/v850.md (addsi3_set_flags): New pattern.
13212 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
13213 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
13214 (zero_extendhisi2_v850_set_flags): Likewise.
13215 (zero_extendqisi2_v850_set_flags): Likewise.
13216 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
13217 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
13218 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
13220 * config/v850/v850-protos.h (notice_update_cc): Remove.
13221 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
13222 (v850_print_operand): Handle 'D' and "d".
13223 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
13224 Add handling of arithmetic/logical operations compared against zero.
13225 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
13226 Do not look at v850_compare_op, instead get mode from last argument.
13227 (v850_gen_compare): Remove
13228 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
13229 after reload for prologue insns.
13230 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
13232 (construct_save_jarl): Likewise.
13233 (TARGET_FLAGS_REGNUM): Define.
13234 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
13235 (NOTICE_UPDATE_CC): Remove.
13236 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
13237 than cc0. Conditionalize on reload_completed.
13238 (cmpsi_insn, setfcc_insn): Likewise.
13239 (tst1 splitter): Turn into define_and_split which sets the flags
13241 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
13242 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
13243 (cstoresf4, cstoredf4): Clobber the flags.
13244 (cmpsi, cmpsf, cmpdf): Remove expanders.
13245 (setf_insn): Remove pattern.
13246 (addsi3): Turn into define_and_split which clobbers the flags after
13247 reload and a suitable pattern (addsi3_clobber_flags) for use after
13249 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
13250 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
13251 (ashrsi3, ashrsi3_v850e2): Likewise.
13252 (bins): Clobber the flags.
13253 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
13254 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
13255 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
13256 (call_value_internal_short, call_value_internal_long): Likewise.
13257 (callt_save_interrupt, callt_return_interrupt): Likewise.
13258 (save_interrupt, return_interrupt): Likewise.
13259 (callt_save_all_interrupt, save_all_interrupt): Likewise.
13260 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
13261 (restore_all_interrupt, _restore_all_interrupt): Likewise.
13262 (All FP comparisons): Only allow after reload has completed.
13264 (divh, divhu): Tweak output template.
13265 (branch_z_normal, branch_z_invert): Remove
13266 (branch_nz_normal, branch_nz_invert): Likewise.
13267 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
13269 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
13270 * config/v850/v850.c (notice_update_cc): Remove.
13271 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
13272 (CC_NO_CARRY): Likewise.
13273 (NOTICE_UPDATE_CC): Define to nothing.
13274 * config/v850/v850.md: Remove block comment on cc0 handling
13275 Remove "cc" attribute from all patterns. Remove cc_status handling
13276 from all patterns. Minor formatting fixes.
13278 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13280 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
13281 (cortex-a76.cortex-a55): Likewise.
13282 * config/aarch64/aarch64-tune.md: Regenerate.
13283 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
13284 cortex-a76.cortex-a55.
13286 2018-06-27 Jeff Law <law@redhat.com>
13288 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
13289 (MULTILIB_DIRNAMES): Similarly.
13291 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
13293 * gimple.h (gimple_return_retbnd): Delete.
13294 (gimple_return_set_retbnd): Likewise.
13295 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
13296 gimple_return_set_retbnd.
13297 * gimple-pretty-print.c (dump_gimple_return): Remove call to
13298 gimple_return_retbnd and adjust.
13299 * tree-inline.h (struct copy_body_data): Remove retbnd field.
13300 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
13301 Explicitly return NULL in a couple more cases. Move assertion
13302 on debug statements and remove unreachable code.
13303 (reset_debug_binding): Do not test id->retbnd.
13304 (expand_call_inline): Do not set it.
13306 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
13308 * configure.ac: Add --disable-gcov option.
13309 * configure: Regenerate.
13310 * Makefile.in: Honour @enable_gcov@.
13311 * doc/install.texi: Document --disable-gcov.
13313 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13315 * config/arm/arm-cpus.in (cortex-a76): New entry.
13316 (cortex-a76.cortex-a55): Likewise.
13317 * config/arm/arm-tables.opt: Regenerate.
13318 * config/arm/arm-tune.md: Likewise.
13319 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
13320 * doc/invoke.texi (ARM Options): Document cortex-a76 and
13321 cortex-a76.cortex-a55.
13323 2018-06-27 Tamar Christina <tamar.christina@arm.com>
13326 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
13328 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
13330 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
13332 (EPILOGUE_USES): Likewise.
13334 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
13336 * tree-inline.c (remap_location): New function extracted from...
13337 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
13338 (copy_phis_for_bb): ...here. Call remap_location.
13339 (copy_cfg_body): Adjust call to copy_edges_for_bb.
13341 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
13343 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
13344 unaligned vsx for 16B memset.
13346 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
13349 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
13350 ieee128_float_type_node to long_double_type_node unless
13351 TARGET_LONG_DOUBLE_128 is set.
13353 2018-06-26 David Malcolm <dmalcolm@redhat.com>
13355 * cfgloop.c (get_loop_location): Convert return type from
13356 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
13357 by implicit construction from rtx_insn *, and using
13358 dump_user_location_t::from_function_decl for the fallback case.
13359 * cfgloop.h (get_loop_location): Convert return type from
13360 location_t to dump_user_location_t.
13361 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
13362 dump_printf_loc to pass in a dump_location_t rather than a
13363 location_t, via the gimple stmt.
13364 * coverage.c (get_coverage_counts): Update calls to
13365 dump_printf_loc to pass in dump_location_t rather than a
13367 * doc/optinfo.texi (Dump types): Convert example of
13368 dump_printf_loc from taking "locus" to taking "insn". Update
13369 description of the "_loc" calls to cover dump_location_t.
13370 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
13372 (dump_user_location_t::dump_user_location_t): New constructors,
13373 from gimple *stmt and rtx_insn *.
13374 (dump_user_location_t::from_function_decl): New function.
13375 (dump_loc): Make static.
13376 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
13377 const dump_location_t &.
13378 (dump_generic_expr_loc): Delete.
13379 (dump_printf_loc): Convert param "loc" from location_t to
13380 const dump_location_t &.
13381 (selftest::test_impl_location): New function.
13382 (selftest::dumpfile_c_tests): New function.
13383 * dumpfile.h: Include "profile-count.h".
13384 (class dump_user_location_t): New class.
13385 (struct dump_impl_location_t): New struct.
13386 (class dump_location_t): New class.
13387 (dump_printf_loc): Convert 2nd param from source_location to
13388 const dump_location_t &.
13389 (dump_generic_expr_loc): Delete.
13390 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
13391 const dump_location_t &.
13392 * gimple-fold.c (fold_gimple_assign): Update call to
13393 dump_printf_loc to pass in a dump_location_t rather than a
13394 location_t, via the gimple stmt.
13395 (gimple_fold_call): Likewise.
13396 * gimple-loop-interchange.cc
13397 (loop_cand::analyze_iloop_reduction_var): Update for change to
13398 check_reduction_path.
13399 (tree_loop_interchange::interchange): Update for change to
13400 find_loop_location.
13401 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
13402 change in return-type of find_loop_location.
13403 (graphite_regenerate_ast_isl): Likewise.
13404 * graphite-optimize-isl.c (optimize_isl): Likewise.
13405 * graphite.c (graphite_transform_loops): Likewise.
13406 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
13407 pass in a dump_location_t rather than a location_t, via the
13409 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13410 * ipa.c (walk_polymorphic_call_targets): Likewise.
13411 * loop-unroll.c (report_unroll): Convert "locus" param from
13412 location_t to dump_location_t.
13413 (decide_unrolling): Update for change to get_loop_location's
13415 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
13416 location_t to dump_user_location_t.
13417 (grid_find_single_omp_among_assignments_1): Updates calls to
13418 dump_printf_loc to pass in a dump_location_t rather than a
13419 location_t, via the gimple stmt.
13420 (grid_parallel_clauses_gridifiable): Convert "tloc" from
13421 location_t to dump_location_t. Updates calls to dump_printf_loc
13422 to pass in a dump_location_t rather than a location_t, via the
13424 (grid_inner_loop_gridifiable_p): Likewise.
13425 (grid_dist_follows_simple_pattern): Likewise.
13426 (grid_gfor_follows_tiling_pattern): Likewise.
13427 (grid_target_follows_gridifiable_pattern): Likewise.
13428 (grid_attempt_target_gridification): Convert initialization
13429 of local "grid" from memset to zero-initialization; FIXME: does
13430 this require C++11? Update call to dump_printf_loc to pass in a
13431 optinfo_location rather than a location_t, via the gimple stmt.
13432 * profile.c (read_profile_edge_counts): Updates call to
13433 dump_printf_loc to pass in a dump_location_t rather than a
13435 (compute_branch_probabilities): Likewise.
13436 * selftest-run-tests.c (selftest::run_tests): Call
13438 * selftest.h (dumpfile_c_tests): New decl.
13439 * tree-loop-distribution.c (pass_loop_distribution::execute):
13440 Update for change in return type of find_loop_location.
13441 * tree-parloops.c (parallelize_loops): Likewise.
13442 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
13443 "locus" from location_t to dump_user_location_t.
13444 (canonicalize_loop_induction_variables): Likewise.
13445 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
13446 for change in return type of find_loop_location.
13447 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
13448 to dump_printf_loc to pass in a dump_location_t rather than a
13449 location_t, via the stmt.
13450 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
13452 * tree-vect-loop-manip.c (find_loop_location): Convert return
13453 type from source_location to dump_user_location_t.
13454 (vect_do_peeling): Update for above change.
13455 (vect_loop_versioning): Update for change in type of
13457 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
13458 from location_t to dump_user_location_t.
13459 (vect_estimate_min_profitable_iters): Update for change in type
13461 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
13462 location_t to dump_location_t.
13463 (vect_slp_bb): Update for change in type of vect_location.
13464 * tree-vectorizer.c (vect_location): Convert from source_location
13465 to dump_user_location_t.
13466 (try_vectorize_loop_1): Update for change in vect_location's type.
13467 (vectorize_loops): Likewise.
13468 (increase_alignment): Likewise.
13469 * tree-vectorizer.h (vect_location): Convert from source_location
13470 to dump_user_location_t.
13471 (find_loop_location): Convert return type from source_location to
13472 dump_user_location_t.
13473 (check_reduction_path): Convert 1st param from location_t to
13474 dump_user_location_t.
13475 * value-prof.c (check_counter): Update call to dump_printf_loc to
13476 pass in a dump_user_location_t rather than a location_t; update
13477 call to error_at for change in type of "locus".
13478 (check_ic_target): Update call to dump_printf_loc to
13479 pass in a dump_user_location_t rather than a location_t, via the
13482 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
13484 * config/s390/s390.h (enum processor_flags): Do not use
13485 default tune parameter when -march was specified.
13487 2018-06-26 Jakub Jelinek <jakub@redhat.com>
13490 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
13491 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
13494 2018-06-26 Richard Biener <rguenther@suse.de>
13496 PR tree-optimization/86287
13498 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
13499 (vect_analyze_loop): Initialize n_stmts.
13501 2018-06-26 Richard Biener <rguenther@suse.de>
13503 PR middle-end/86271
13504 * fold-const.c (fold_convertible_p): Pointer extension
13507 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
13510 * dwarf2out.c (loc_list_has_views): Adjust comments.
13511 (dw_loc_list): Split single cross-partition range with
13514 2018-06-25 Jeff Law <law@redhat.com>
13516 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
13517 on -mbig-switch by default.
13519 * config/v850/predicates.md (const_float_1_operand): Fix match_code
13521 (const_float_0_operand): Remove unused predicate.
13522 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
13523 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
13524 (recipsf2): New expander. Original pattern now called
13526 (recipdf2, recipdf2_insn): Similarly.
13527 (rsqrtsf2, rsqrtsf2_insn): Similarly
13528 (rsqrtdf2, rsqrtdf2_insn): Similarly
13530 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
13532 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
13533 Simplify logic for FreeBSD (twice).
13535 2018-06-25 Martin Sebor <msebor@redhat.com>
13537 PR tree-optimization/86204
13538 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
13539 a strnlen result if it's less than the length of the string.
13541 2018-06-25 Martin Sebor <msebor@redhat.com>
13543 PR tree-optimization/85700
13544 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
13545 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
13546 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
13548 2018-06-25 Martin Sebor <msebor@redhat.com>
13550 * doc/extend.texi (Zero-length arrays): Update and clarify.
13552 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
13554 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
13555 added IEEE/IBM long double multilib support on PowerPC little
13556 endian Linux systems.
13557 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
13558 (MULTILIB_DEFAULTS): Likewise.
13559 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13561 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
13562 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
13563 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
13565 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
13567 PR middle-end/86311
13568 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
13569 (REORDER_45): Likewise.
13571 2018-06-25 Jeff Law <law@redhat.com>
13573 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
13574 dividend to 32 bits. Adjust length.
13575 (udivmodhi4): Cleanup output template. Fix length.
13577 2018-06-25 Carl Love <cel@us.ibm.com>
13579 * config/rs6000/vsx.md: Change word selector to prefered location.
13581 2018-06-25 Richard Biener <rguenther@suse.de>
13583 PR tree-optimization/86304
13584 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
13585 epilogue-if-converted loops as well.
13587 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
13589 * lto-section-out.c (lto_begin_section): Do not print section
13590 name for noaddr and unnumbered dumps.
13592 2018-06-25 Richard Biener <rguenther@suse.de>
13594 * tree-vectorizer.h (struct vec_info_shared): New structure
13595 with parts split out from struct vec_info and loop_nest from
13596 struct _loop_vec_info.
13597 (struct vec_info): Adjust accordingly.
13598 (struct _loop_vec_info): Likewise.
13599 (LOOP_VINFO_LOOP_NEST): Adjust.
13600 (LOOP_VINFO_DATAREFS): Likewise.
13601 (LOOP_VINFO_DDRS): Likewise.
13602 (struct _bb_vec_info): Likewise.
13603 (BB_VINFO_DATAREFS): Likewise.
13604 (BB_VINFO_DDRS): Likewise.
13605 (struct _stmt_vec_info): Add dr_aux member.
13606 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
13607 (DR_MISALIGNMENT_UNINITIALIZED): New.
13608 (set_dr_misalignment): Adjust.
13609 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
13610 (vect_analyze_loop): Adjust prototype.
13611 (vect_analyze_loop_form): Likewise.
13612 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
13613 Compute dependences lazily.
13614 (vect_record_base_alignments): Use shared datarefs/ddrs.
13615 (vect_verify_datarefs_alignment): Likewise.
13616 (vect_analyze_data_refs_alignment): Likewise.
13617 (vect_analyze_data_ref_accesses): Likewise.
13618 (vect_analyze_data_refs): Likewise.
13619 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
13620 constructor parameter for shared part.
13621 (vect_analyze_loop_form): Pass in shared part and adjust.
13622 (vect_analyze_loop_2): Pass in storage for the number of
13623 stmts. Move loop nest finding to the caller. Compute
13625 (vect_analyze_loop): Pass in shared part.
13626 (vect_transform_loop): Verify shared datarefs are unchanged.
13627 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
13628 constructor parameter for shared part.
13629 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
13630 (vect_slp_bb): Verify shared datarefs are unchanged before
13632 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
13634 (new_stmt_vec_info): Initialize DR_AUX misalignment to
13635 DR_MISALIGNMENT_UNINITIALIZED.
13636 * tree-vectorizer.c (vec_info::vec_info): Add constructor
13637 parameter for shared part.
13638 (vec_info::~vec_info): Adjust.
13639 (vec_info_shared::vec_info_shared): New.
13640 (vec_info_shared::~vec_info_shared): Likewise.
13641 (vec_info_shared::save_datarefs): Likewise.
13642 (vec_info_shared::check_datarefs): Likewise.
13643 (try_vectorize_loop_1): Construct shared part live for analyses
13644 of a single loop for multiple vector sizes.
13645 * tree-parloops.c (gather_scalar_reductions): Adjust.
13647 2018-06-25 Richard Biener <rguenther@suse.de>
13649 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
13650 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
13651 (vect_analyze_data_refs): Remove similar code from here and
13652 simplify accordingly.
13654 2018-06-25 Richard Biener <rguenther@suse.de>
13656 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
13657 for reverse storage order accesses rather than asserting
13658 they cannot happen here.
13660 2018-06-25 Tom de Vries <tdevries@suse.de>
13663 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
13664 Use data16 instead of .byte for insn prefix.
13666 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
13669 * parser.c (make_char_string_pack): Pass this literal chars
13670 through cpp_interpret_string.
13671 (cp_parser_userdef_numeric_literal): Check the result of
13672 make_char_string_pack.
13674 2018-06-24 Maya Rashish <coypu@sdf.org>
13676 * ginclude/stddef.h: Simplify conditions around avoiding
13677 re-definition of __size_t.
13679 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
13681 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
13682 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
13684 2018-06-22 Maya Rashish <coypu@sdf.org>
13686 * doc/invoke.texi (mno-fancy-math-387): Update for changes
13687 made to OpenBSD and NetBSD through the years.
13689 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
13691 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
13692 behavior of vec_pack (vector double, vector double) to match
13693 behavior of vec_float2 (vector double, vector double).
13695 2018-06-22 Olivier Hainque <hainque@adacore.com>
13697 * gimplify.c (gimplify_function_tree): Prevent creation
13698 of a trampoline for the address of the current function
13699 passed to entry/exit instrumentation hooks.
13701 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
13704 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
13707 2018-06-22 Martin Liska <mliska@suse.cz>
13709 PR tree-optimization/86263
13710 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
13711 Bail out if is_enabled is false.
13712 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
13714 (jump_table_cluster::is_enabled): New function.
13716 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
13718 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
13719 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
13720 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
13721 (lto_input_ts_binfo_tree_pointers): Likewise.
13722 * tree-streamer-out.c (streamer_write_tree_bitfields,
13723 write_ts_binfo_tree_pointers): Likewise.
13724 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
13726 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
13728 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
13730 2018-06-22 Martin Liska <mliska@suse.cz>
13732 * symbol-summary.h (get): Make it pure and inline move
13733 functionality from ::get function.
13734 (get): Remove and inline into ::get and ::get_create.
13735 (get_create): Move code from ::get function.
13737 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13740 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
13741 -x assembler-with-cpp.
13743 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13745 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
13746 _FILE_OFFSET_BITS=64 for C++.
13748 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
13750 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
13751 conversion insn that shows up when pr85657-3.c is compiled using
13752 IEEE 128-bit long double.
13753 (neg<mode>2_internal): Use the correct mode to check whether the
13754 mode is IBM extended.
13755 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
13756 multiply and divide external functions from being created more
13759 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
13761 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
13763 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
13764 the edge can be forwarded.
13765 (cfg_layout_merge_blocks): Likewise.
13767 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
13769 * except.c (finish_eh_generation): Commit edge insertions only after
13770 the EH edges have been redirected from post-landing to landing pads.
13772 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
13774 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
13775 create_tmp_var_for to create the FRAME decl.
13776 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
13778 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
13780 * tree-inline.c (copy_edges_for_bb): Minor tweak.
13781 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
13782 debug statement when resetting its value.
13783 (expand_call_inline): Copy the locus of the call onto the assignment
13784 of the return value, if any. Use local variable in more cases.
13786 2018-06-21 Martin Liska <mliska@suse.cz>
13788 * ipa-pure-const.c (propagate_nothrow): Use
13789 funct_state_summaries->get.
13790 (dump_malloc_lattice): Likewise.
13791 (propagate_malloc): Likewise.
13793 2018-06-21 Richard Biener <rguenther@suse.de>
13795 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
13796 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
13797 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
13799 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
13800 BLOCK_ABSTRACT_ORIGIN unconditionally.
13802 2018-06-21 David Malcolm <dmalcolm@redhat.com>
13804 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
13806 * ipa-reference.c (ipa_reference_c_finalize): Delete
13807 ipa_ref_opt_sum_summaries and set it to NULL.
13809 2018-06-21 Tom de Vries <tdevries@suse.de>
13811 PR tree-optimization/85859
13812 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
13813 test with comment from bb_no_side_effects_p.
13815 2018-06-21 Richard Biener <rguenther@suse.de>
13817 PR tree-optimization/86232
13818 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
13819 max for constant niter.
13821 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
13823 * config/aarch64/aarch64-simd.md
13824 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
13826 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
13828 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
13829 Make opernads of the unspec commutative.
13831 2018-06-21 Richard Biener <rguenther@suse.de>
13833 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
13834 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
13835 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
13836 (vect_analyze_data_ref_dependence): Re-order checks to deal with
13838 (vect_record_base_alignments): Do not record base alignment
13839 for gathers or scatters.
13840 (vect_compute_data_ref_alignment): Drop return value that is always
13841 true. Bail out early for gathers or scatters.
13842 (vect_enhance_data_refs_alignment): Bail out early for gathers
13844 (vect_find_same_alignment_drs): Likewise.
13845 (vect_analyze_data_refs_alignment): Remove dead code.
13846 (vect_slp_analyze_and_verify_node_alignment): Likewise.
13847 (vect_analyze_data_refs): For possible gathers or scatters do
13848 not create an alternate DR, just check their possible validity
13849 and mark them. Adjust DECL_NONALIASED handling to not rely
13850 on DR_BASE_ADDRESS.
13851 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
13852 update inits of gathers or scatters.
13853 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
13854 Also copy gather/scatter flag to pattern vinfo.
13856 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
13858 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
13859 behavior of vec_packsu (vector unsigned long long, vector unsigned
13860 long long) to match behavior of vec_packs with same signature.
13862 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
13863 Thomas Schwinge <thomas@codesourcery.com>
13864 Cesar Philippidis <cesar@codesourcery.com>
13866 * gimplify.c (gimplify_scan_omp_clauses): Add support for
13867 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
13868 (gimplify_adjust_omp_clauses): Likewise.
13869 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
13870 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
13871 (gimplify_omp_target_update): Update handling of acc update and
13873 * omp-low.c (install_var_field): Remove unused parameter
13874 base_pointers_restrict.
13875 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
13876 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
13878 (omp_target_base_pointers_restrict_p): Delete.
13879 (scan_omp_target): Update call to scan_sharing_clauses.
13880 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
13882 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
13883 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
13884 (convert_local_omp_clauses): Likewise.
13885 * tree-pretty-print.c (dump_omp_clause): Likewise.
13886 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
13888 (omp_clause_code_name): Likewise.
13890 2018-06-20 Jakub Jelinek <jakub@redhat.com>
13893 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
13896 PR tree-optimization/86231
13897 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
13898 anti-range don't overwrite *vr0min before using it to compute *vr0max.
13900 2018-06-20 Tom de Vries <tdevries@suse.de>
13902 PR tree-optimization/86097
13903 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
13904 iv type if signedness of iv type is not the same as that of *nit.
13906 2018-06-20 Jakub Jelinek <jakub@redhat.com>
13908 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
13909 EDGE_EH edges, verify they are all EDGE_EH.
13911 2018-06-20 Maya Rashish <coypu@sdf.org>
13913 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
13915 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13917 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
13918 * config/aarch64/aarch64.c (xgene1_tunings): Add
13919 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
13920 (aarch64_mode_valid_for_sched_fusion_p):
13921 Allow 16-byte modes.
13922 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
13923 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
13925 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
13927 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
13928 * config/aarch64/iterators.md (VQ2): New mode iterator.
13930 2018-06-20 Martin Liska <mliska@suse.cz>
13932 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
13933 Change default ratio from 10 to 8.
13935 2018-06-20 Martin Liska <mliska@suse.cz>
13937 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
13939 (bit_test_cluster::find_bit_tests): Likewise.
13940 (switch_decision_tree::analyze_switch_statement): Find clusters.
13941 * tree-switch-conversion.h (struct jump_table_cluster): Document
13944 2018-06-20 Martin Liska <mliska@suse.cz>
13946 * tree-switch-conversion.c (switch_conversion::collect):
13947 Record m_uniq property.
13948 (switch_conversion::expand): Bail out for special conditions.
13949 (group_cluster::~group_cluster): New.
13950 (group_cluster::group_cluster): Likewise.
13951 (group_cluster::dump): Likewise.
13952 (jump_table_cluster::emit): New.
13953 (switch_decision_tree::fix_phi_operands_for_edges): New.
13954 (struct case_node): Remove struct.
13955 (jump_table_cluster::can_be_handled): New.
13956 (case_values_threshold): Moved to header.
13957 (reset_out_edges_aux): Likewise.
13958 (jump_table_cluster::is_beneficial): New.
13959 (bit_test_cluster::can_be_handled): Likewise.
13960 (add_case_node): Remove.
13961 (bit_test_cluster::is_beneficial): New.
13962 (case_bit_test::cmp): New.
13963 (bit_test_cluster::emit): New.
13964 (expand_switch_as_decision_tree_p): Remove.
13965 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
13966 (fix_phi_operands_for_edge): Likewise.
13967 (switch_decision_tree::analyze_switch_statement): New.
13968 (compute_cases_per_edge): Move ...
13969 (switch_decision_tree::compute_cases_per_edge): ... here.
13970 (try_switch_expansion): Likewise.
13971 (switch_decision_tree::try_switch_expansion): Likewise.
13972 (record_phi_operand_mapping): Likewise.
13973 (switch_decision_tree::record_phi_operand_mapping): Likewise.
13974 (emit_case_decision_tree): Likewise.
13975 (switch_decision_tree::emit): Likewise.
13976 (balance_case_nodes): Likewise.
13977 (switch_decision_tree::balance_case_nodes): Likewise.
13978 (dump_case_nodes): Likewise.
13979 (switch_decision_tree::dump_case_nodes): Likewise.
13980 (emit_jump): Likewise.
13981 (switch_decision_tree::emit_jump): Likewise.
13982 (emit_cmp_and_jump_insns): Likewise.
13983 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
13984 (emit_case_nodes): Likewise.
13985 (switch_decision_tree::emit_case_nodes): Likewise.
13986 (conditional_probability): Remove.
13987 * tree-switch-conversion.h (enum cluster_type): New.
13989 (struct cluster): Likewise.
13990 (cluster::cluster): Likewise.
13991 (struct simple_cluster): Likewise.
13992 (simple_cluster::simple_cluster): Likewise.
13993 (struct group_cluster): Likewise.
13994 (struct jump_table_cluster): Likewise.
13995 (struct bit_test_cluster): Likewise.
13996 (struct min_cluster_item): Likewise.
13997 (struct case_tree_node): Likewise.
13998 (case_tree_node::case_tree_node): Likewise.
13999 (jump_table_cluster::case_values_threshold): Likewise.
14000 (struct case_bit_test): Likewise.
14001 (struct switch_decision_tree): Likewise.
14002 (struct switch_conversion): Likewise.
14003 (switch_decision_tree::reset_out_edges_aux): Likewise.
14005 2018-06-20 Martin Liska <mliska@suse.cz>
14007 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
14008 (hoist_edge_and_branch_if_true): Likewise.
14009 (expand_switch_using_bit_tests_p): Likewise.
14010 (struct case_bit_test): Likewise.
14011 (case_bit_test_cmp): Likewise.
14012 (emit_case_bit_tests): Likewise.
14013 (switch_conversion::switch_conversion): New class.
14014 (struct switch_conv_info): Remove old struct.
14015 (collect_switch_conv_info): More to ...
14016 (switch_conversion::collect): ... this.
14017 (check_range): Likewise.
14018 (switch_conversion::check_range): Likewise.
14019 (check_all_empty_except_final): Likewise.
14020 (switch_conversion::check_all_empty_except_final): Likewise.
14021 (check_final_bb): Likewise.
14022 (switch_conversion::check_final_bb): Likewise.
14023 (create_temp_arrays): Likewise.
14024 (switch_conversion::create_temp_arrays): Likewise.
14025 (free_temp_arrays): Likewise.
14026 (gather_default_values): Likewise.
14027 (switch_conversion::gather_default_values): Likewise.
14028 (build_constructors): Likewise.
14029 (switch_conversion::build_constructors): Likewise.
14030 (constructor_contains_same_values_p): Likewise.
14031 (switch_conversion::contains_same_values_p): Likewise.
14032 (array_value_type): Likewise.
14033 (switch_conversion::array_value_type): Likewise.
14034 (build_one_array): Likewise.
14035 (switch_conversion::build_one_array): Likewise.
14036 (build_arrays): Likewise.
14037 (switch_conversion::build_arrays): Likewise.
14038 (gen_def_assigns): Likewise.
14039 (switch_conversion::gen_def_assigns): Likewise.
14040 (prune_bbs): Likewise.
14041 (switch_conversion::prune_bbs): Likewise.
14042 (fix_phi_nodes): Likewise.
14043 (switch_conversion::fix_phi_nodes): Likewise.
14044 (gen_inbound_check): Likewise.
14045 (switch_conversion::gen_inbound_check): Likewise.
14046 (process_switch): Use the newly created class.
14047 (switch_conversion::expand): New.
14048 (switch_conversion::~switch_conversion): New.
14049 * tree-switch-conversion.h: New file.
14051 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
14053 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
14054 tree-vect-patterns.c.
14055 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
14056 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
14057 (vect_recog_sad_pattern): Likewise.
14058 (vect_recog_widen_sum_pattern): Likewise.
14059 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
14060 (vect_recog_widen_shift_pattern): Remove the type_in argument.
14061 (vect_recog_rotate_pattern): Likewise.
14062 (vect_recog_mult_pattern): Likewise.
14063 (vect_recog_vector_vector_shift_pattern): Likewise.
14064 (vect_recog_divmod_pattern): Likewise.
14065 (vect_recog_mixed_size_cond_pattern): Likewise.
14066 (vect_recog_bool_pattern): Likewise.
14067 (vect_recog_mask_conversion_pattern): Likewise.
14068 (vect_try_gather_scatter_pattern): Likewise.
14069 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
14070 (vect_recog_over_widening_pattern): Likewise.
14071 (vect_recog_gather_scatter_pattern): Likewise.
14072 (vect_recog_func_ptr): Move from tree-vectorizer.h
14073 (vect_vect_recog_func_ptrs): Move further down the file.
14074 (vect_recog_func): Likewise. Remove the third argument.
14075 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
14076 (vect_pattern_recog_1): Expect the pattern function to do any
14077 necessary target tests. Also expect it to provide a vector type.
14078 Remove the type_in handling.
14080 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
14082 * tree-vect-patterns.c (vect_pattern_detected): New function.
14083 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
14084 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
14085 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
14086 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
14087 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
14088 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
14089 (vect_recog_mask_conversion_pattern)
14090 (vect_try_gather_scatter_pattern): Likewise.
14092 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
14094 * tree-vect-patterns.c (vect_get_internal_def): New function.
14095 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14096 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
14097 (search_type_for_mask_1): Use it.
14099 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
14101 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
14102 redundant WIDEN_SUM_EXPR handling.
14103 (vect_recog_sad_pattern): Likewise.
14105 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
14107 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
14108 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
14109 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
14110 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
14111 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
14113 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
14115 * tree-vect-stmts.c (vectorizable_call): Make sure that we
14116 use the stmt_vec_info of the original bb statement for the
14117 new zero assignment, even if the call is part of a pattern.
14119 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
14121 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
14122 that the sequence is attached to the original statement rather
14123 than the pattern statement.
14124 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
14125 PATTERN_DEF_SEQ from the original statement rather than
14126 the main pattern statement.
14127 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
14128 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
14129 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
14131 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
14133 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
14134 definition statements before the early exit for statements that aren't
14136 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
14138 (vect_transform_loop): ...here. Process pattern definition
14139 statements without first checking whether the main pattern
14140 statement is live or relevant.
14142 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
14144 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
14145 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
14147 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
14149 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
14150 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
14151 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
14152 (expand_block_compare): Change select_block_compare_mode call.
14153 (expand_strncmp_align_check): Use new functions, fix comment.
14154 (emit_final_str_compare_gpr): New function.
14155 (expand_strn_compare): Refactor and clean up code.
14156 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
14158 2018-06-19 Tony Reix <tony.reix@atos.com>
14159 Damien Bergamini <damien.bergamini@atos.com>
14160 David Edelsohn <dje.gcc@gmail.com>
14162 * collect2.c (static_obj): New variable.
14163 (static_libs): New variable.
14164 (is_in_list): Uncomment declaration.
14165 (main): Track AIX libraries linked statically.
14166 (is_in_list): Uncomment definition.
14167 (scan_prog_file): Don't add AIX shared libraries initializer
14168 to constructor list if linking statically.
14170 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
14172 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
14174 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
14176 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
14178 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
14181 2018-06-19 Martin Liska <mliska@suse.cz>
14183 * config/i386/i386.c (ix86_can_inline_p): Do not use
14184 ipa_fn_summaries::get_create.
14185 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
14187 (devirtualization_time_bonus): Likewise.
14188 (ipcp_propagate_stage): Likewise.
14189 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
14190 (edge_set_predicate): Likewise.
14191 (evaluate_conditions_for_known_args): Likewise.
14192 (evaluate_properties_for_edge): Likewise.
14193 (ipa_call_summary::reset): Tranform to ...
14194 (ipa_call_summary::~ipa_call_summary): ... this.
14195 (ipa_fn_summary::reset): Transform to ...
14196 (ipa_fn_summary::~ipa_fn_summary): ... this.
14197 (ipa_fn_summary_t::remove): Rename to ...
14198 (ipa_fn_summary_t::remove_callees): ... this.
14199 (ipa_fn_summary_t::duplicate): Use placement new
14200 instead of memory copy.
14201 (ipa_call_summary_t::duplicate): Likewise.
14202 (ipa_call_summary_t::remove): Remove.
14203 (dump_ipa_call_summary): Change get_create to get.
14204 (ipa_dump_fn_summary): Dump only when summary exists.
14205 (analyze_function_body): Use symbol_summary::get instead
14207 (compute_fn_summary): Likewise.
14208 (estimate_edge_devirt_benefit): Likewise.
14209 (estimate_edge_size_and_time): Likewise.
14210 (inline_update_callee_summaries): Likewise.
14211 (remap_edge_change_prob): Likewise.
14212 (remap_edge_summaries): Likewise.
14213 (ipa_merge_fn_summary_after_inlining): Likewise.
14214 (write_ipa_call_summary): Likewise.
14215 (ipa_fn_summary_write): Likewise.
14216 (ipa_free_fn_summary): Likewise.
14217 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
14218 (struct ipa_call_summary): Likewise.
14219 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
14221 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
14222 (estimate_size_after_inlining): Likewise.
14223 (estimate_growth): Likewise.
14224 (growth_likely_positive): Likewise.
14225 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
14226 (inline_call): Likewise.
14227 * ipa-inline.c (caller_growth_limits): Likewise.
14228 (can_inline_edge_p): Likewise.
14229 (can_inline_edge_by_limits_p): Likewise.
14230 (compute_uninlined_call_time): Likewise.
14231 (compute_inlined_call_time): Likewise.
14232 (want_inline_small_function_p): Likewise.
14233 (edge_badness): Likewise.
14234 (update_caller_keys): Likewise.
14235 (update_callee_keys): Likewise.
14236 (inline_small_functions): Likewise.
14237 (inline_to_all_callers_1): Likewise.
14238 (dump_overall_stats): Likewise.
14239 (early_inline_small_functions): Likewise.
14240 (early_inliner): Likewise.
14241 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
14242 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14243 * ipa-pure-const.c (malloc_candidate_p): Likewise.
14244 * ipa-split.c (execute_split_functions): Likewise.
14245 * symbol-summary.h: Likewise.
14246 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
14248 2018-06-19 Richard Biener <rguenther@suse.de>
14250 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
14251 (vectorize_loops): ... here. Fix dbgcnt handling.
14252 (try_vectorize_loop): Wrap try_vectorize_loop_1.
14254 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
14257 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
14258 ieee128 argument takes up only one (vector) register, not two (floating
14261 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
14263 * gimplify.c (gimplify_init_constructor): Really never clear for an
14264 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
14266 2018-06-19 Richard Biener <rguenther@suse.de>
14268 PR tree-optimization/86179
14269 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
14270 after failed recognition.
14272 2018-06-18 Martin Sebor <msebor@redhat.com>
14274 PR middle-end/85602
14275 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
14276 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
14277 Handle integer subtraction.
14278 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
14279 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
14281 2018-06-18 David Malcolm <dmalcolm@redhat.com>
14283 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
14284 param from rtx to rtx_insn *.
14285 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
14287 (frv_ifcvt_modify_insn): Likwise.
14288 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
14289 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
14290 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
14291 as_a <rtx_insn *> cast to local "unprotected_region" once
14292 it's been established that it's not NULL or pc_rtx.
14293 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
14294 param "sethi" from rtx to rtx_insn *.
14295 (nds32_group_float_insns): Likewise for param "insn".
14296 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
14298 (vax_output_int_subtract): Likewise.
14299 * config/vax/vax.c (vax_output_int_add): Likewise for param
14301 (vax_output_int_subtract): Likewise.
14302 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
14303 (emit_pattern_after): Likewise for param "after".
14304 (emit_insn_after): Likewise.
14305 (emit_jump_insn_after): Likewise.
14306 (emit_call_insn_after): Likewise.
14307 (emit_debug_insn_after): Likewise.
14308 (emit_pattern_before): Likewise for param "before".
14309 (emit_insn_before): Likewise.
14310 (emit_jump_insn_before): Likewise.
14311 * final.c (get_insn_template): Likewise for param "insn", removing
14313 * output.h (get_insn_template): Likewise for 2nd param.
14314 * rtl.h (emit_insn_before): Likewise.
14315 (emit_jump_insn_before): Likewise.
14316 (emit_debug_insn_before_noloc): Likewise.
14317 (emit_insn_after): Likewise.
14318 (emit_jump_insn_after): Likewise.
14319 (emit_call_insn_after): Likewise.
14320 (emit_debug_insn_after): Likewise.
14321 (set_insn_deleted): Likewise for param.
14323 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
14326 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
14327 floating point modes, so that IFmode is numerically greater than
14328 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
14329 to declare the ordering. This prevents IFmode from being
14330 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
14331 machine. Include rs6000-modes.h to share the fractional values
14332 between genmodes* and the rest of the compiler.
14333 (IFmode): Likewise.
14334 (KFmode): Likewise.
14335 (TFmode): Likewise.
14336 * config/rs6000/rs6000-modes.h: New file.
14337 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
14338 meaning of rs6000_long_double_size so that 126..128 selects an
14339 appropriate 128-bit floating point type.
14340 (rs6000_option_override_internal): Likewise.
14341 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
14342 (TARGET_LONG_DOUBLE_128): Change the meaning of
14343 rs6000_long_double_size so that 126..128 selects an appropriate
14344 128-bit floating point type.
14345 (LONG_DOUBLE_TYPE_SIZE): Update comment.
14346 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
14347 source and destination to match the standard usage.
14348 (truncifkf2): Likewise.
14349 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
14350 ISA 2.07 to use an explicit clobber, instead of passing in a
14352 (copysign<mode>3_soft): Likewise.
14354 2018-06-18 David Malcolm <dmalcolm@redhat.com>
14356 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
14357 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
14358 (vect_slp_analyze_instance_dependence): Likewise.
14359 (vect_enhance_data_refs_alignment): Likewise.
14360 (vect_analyze_data_refs_alignment): Likewise.
14361 (vect_slp_analyze_and_verify_instance_alignment
14362 (vect_analyze_data_ref_accesses): Likewise.
14363 (vect_prune_runtime_alias_test_list): Likewise.
14364 (vect_analyze_data_refs): Likewise.
14365 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
14366 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14367 (vect_analyze_scalar_cycles_1): Likewise.
14368 (vect_get_loop_niters): Likewise.
14369 (vect_analyze_loop_form_1): Likewise.
14370 (vect_update_vf_for_slp): Likewise.
14371 (vect_analyze_loop_operations): Likewise.
14372 (vect_analyze_loop): Likewise.
14373 (vectorizable_induction): Likewise.
14374 (vect_transform_loop): Likewise.
14375 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
14376 * tree-vect-slp.c (vect_analyze_slp): Likewise.
14377 (vect_make_slp_decision): Likewise.
14378 (vect_detect_hybrid_slp): Likewise.
14379 (vect_slp_analyze_operations): Likewise.
14380 (vect_slp_bb): Likewise.
14381 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
14382 (vectorizable_bswap): Likewise.
14383 (vectorizable_call): Likewise.
14384 (vectorizable_simd_clone_call): Likewise.
14385 (vectorizable_conversion): Likewise.
14386 (vectorizable_assignment): Likewise.
14387 (vectorizable_shift): Likewise.
14388 (vectorizable_operation): Likewise.
14389 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
14391 2018-06-18 Martin Sebor <msebor@redhat.com>
14393 PR tree-optimization/81384
14394 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
14395 * builtins.c (expand_builtin_strnlen): New function.
14396 (expand_builtin): Call it.
14397 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
14398 * builtins.def (BUILT_IN_STRNLEN): New.
14399 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
14400 Warn for bounds in excess of maximum object size.
14401 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
14402 single-value ranges. Handle strnlen.
14403 (handle_builtin_strlen): Handle strnlen.
14404 (strlen_check_and_optimize_stmt): Same.
14405 * doc/extend.texi (Other Builtins): Document strnlen.
14407 2018-06-18 Maya Rashish <coypu@sdf.org>
14409 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
14410 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
14411 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
14413 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
14415 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
14417 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14419 * tree.c (escaped_string::escape): Replace cast to char * by
14420 const_cast<char *> (unescaped).
14422 2018-06-18 Nick Clifton <nickc@redhat.com>
14425 * tree.c (escaped_string): New class. Converts an unescaped
14426 string into its escaped equivalent.
14427 (warn_deprecated_use): Use the new class to convert the
14428 deprecation message, if present.
14429 (test_escaped_strings): New self test.
14430 (test_c_tests): Add test_escaped_strings.
14431 * doc/extend.texi (deprecated): Add a note that the
14432 deprecation message is affected by the -fmessage-length
14433 option, and that control characters will be escaped.
14434 (#pragma GCC error): Document this pragma.
14435 (#pragma GCC warning): Likewise.
14436 * doc/invoke.texi (-fmessage-length): Document this option's
14437 effect on the #warning and #error preprocessor directives and
14438 the deprecated attribute.
14440 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
14442 * tree.c (decl_value_expr_lookup): Revert latest change.
14443 (decl_value_expr_insert): Likewise.
14445 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
14447 * gimplify.c (nonlocal_vlas): Delete.
14448 (nonlocal_vla_vars): Likewise.
14449 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
14451 (gimplify_body): Do not create and destroy nonlocal_vlas.
14452 * tree-nested.c: Include diagnostic.h.
14453 (use_pointer_in_frame): Tweak.
14454 (lookup_field_for_decl): Add assertion and declare the transformation.
14455 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
14456 internal error when the reference is in a wrong context. Do not
14457 create a debug decl by default.
14458 (note_nonlocal_block_vlas): Delete.
14459 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
14460 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
14461 create a debug decl by default.
14462 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
14463 call is in a wrong context.
14464 (fixup_vla_decls): New function.
14465 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
14466 debug variables were created.
14467 * tree.c (decl_value_expr_lookup): Add checking assertion.
14468 (decl_value_expr_insert): Likewise.
14470 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
14472 PR middle-end/82479
14473 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
14474 * tree-scalar-evolution.c (interpret_expr): Likewise.
14475 (expression_expensive_p): Likewise.
14476 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
14477 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
14478 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
14479 (ssa_defined_by_minus_one_stmt_p): New.
14481 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
14483 PR middle-end/64946
14484 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
14485 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
14486 * dojump.c (do_jump): Likewise.
14487 * expr.c (expand_expr_real_2): Check operand type's sign.
14488 * fold-const.c (const_unop): Handle ABSU_EXPR.
14489 (fold_abs_const): Likewise.
14490 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
14491 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
14492 (strip_sign_op_1): Likesise.
14493 * match.pd: Add new pattern to generate ABSU_EXPR.
14494 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
14495 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
14496 * tree-eh.c (operation_could_trap_helper_p): Likewise.
14497 * tree-inline.c (estimate_operator_cost): Likewise.
14498 * tree-pretty-print.c (dump_generic_node): Likewise.
14499 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
14500 * tree.def (ABSU_EXPR): New.
14502 2018-06-16 Jakub Jelinek <jakub@redhat.com>
14504 PR middle-end/86095
14505 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
14506 documented as preserved for backward compatibility only.
14507 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
14509 PR rtl-optimization/86108
14510 * bb-reorder.c (create_forwarder_block): Renamed to ...
14511 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
14512 jump from new landing pad to the second part.
14513 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
14516 2018-06-15 Jakub Jelinek <jakub@redhat.com>
14518 PR middle-end/85878
14519 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
14520 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
14521 Only call store_expr for halves if the mode is the same.
14523 PR middle-end/86123
14524 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
14525 Fix up comment formatting.
14527 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
14529 * typed-splay-tree.h (typed_splay_tree::remove): New function.
14530 (typed_splay_tree::closure,
14531 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
14532 (typed_splay_tree::typed_splay_tree,
14533 typed_splay_tree::operator =): Declared private.
14534 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
14535 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
14536 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
14537 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
14538 typed_splay_tree::splay_tree_splay,
14539 typed_splay_tree::splay_tree_foreach_helper,
14540 typed_splay_tree::splay_tree_insert,
14541 typed_splay_tree::splay_tree_remove,
14542 typed_splay_tree::splay_tree_lookup,
14543 typed_splay_tree::splay_tree_predecessor,
14544 typed_splay_tree::splay_tree_successor,
14545 typed_splay_tree::splay_tree_min,
14546 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
14547 (typed_splay_tree::root, typed_splay_tree::comp,
14548 typed_splay_tree::delete_key,
14549 typed_splay_tree::delete_value): New data members.
14550 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
14551 typed_splay_tree::remove.
14553 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
14555 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
14556 -mginv and -mno-ginv to the assembler.
14557 * config/mips/mips.opt (-mcrc): New option.
14558 (-mginv): Likewise.
14559 * doc/invoke.text (-mcrc): Document.
14560 (-mginv): Likewise.
14562 2018-06-15 Nick Clifton <nickc@redhat.com>
14565 * tree.c (escaped_string): New class. Converts an unescaped
14566 string into its escaped equivalent.
14567 (warn_deprecated_use): Use the new class to convert the
14568 deprecation message, if present.
14569 (test_escaped_strings): New self test.
14570 (test_c_tests): Add test_escaped_strings.
14571 * doc/extend.texi (deprecated): Add a note that the
14572 deprecation message is affected by the -fmessage-length
14573 option, and that control characters will be escaped.
14574 (#pragma GCC error): Document this pragma.
14575 (#pragma GCC warning): Likewise.
14576 * doc/invoke.texi (-fmessage-length): Document this option's
14577 effect on the #warning and #error preprocessor directives and
14578 the deprecated attribute.
14580 2018-06-15 Richard Biener <rguenther@suse.de>
14582 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
14583 here, also noting vector size used.
14584 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
14585 size used in MSG_OPTIMIZED_LOCATIONS dump.
14586 (pass_slp_vectorize::execute): Adjust.
14588 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
14591 * config/arc/arc.c (arc_return_address_register): Fix
14594 2018-06-15 Richard Biener <rguenther@suse.de>
14596 PR middle-end/86159
14597 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
14598 leave useless conversion stripping to force_gimple_operand_gsi.
14599 (gimplify_build2): Likewise.
14600 (gimplify_build1): Likewise.
14602 2018-06-15 Richard Biener <rguenther@suse.de>
14604 PR middle-end/86076
14605 * tree-cfg.c (move_stmt_op): unshare invariant addresses
14606 before adjusting their block.
14608 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
14610 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
14611 multilibs for *-*-rtems*.
14612 * config/riscv/t-rtems: New file.
14614 2018-06-14 Jakub Jelinek <jakub@redhat.com>
14616 PR middle-end/86122
14617 * match.pd ((A +- CST1) +- CST2): Punt if last resort
14618 unsigned_type_for returns NULL.
14621 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
14622 subregs of multi-word pseudos unless the float mode has word size.
14624 2018-06-14 Richard Biener <rguenther@suse.de>
14626 PR middle-end/86139
14627 * tree-vect-generic.c (build_word_mode_vector_type): Remove
14628 duplicate and harmful type_hash_canon.
14629 * tree.c (type_hash_canon): Assert we didn't find ourselves.
14631 2018-06-14 Richard Biener <rguenther@suse.de>
14634 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
14637 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
14639 * config/rtems.h (STDINT_LONG32): Define.
14641 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
14642 Prachi Godbole <prachi.godbole@imgtec.com>
14644 * config/mips/mips-cpus.def: Define P6600.
14645 * config/mips/mips-tables.opt: Regenerate.
14646 * config/mips/mips.c (mips_ucbranch_type): New enum.
14647 (mips_rtx_cost_data): Add support for P6600.
14648 (mips_issue_rate): Likewise.
14649 (mips_multipass_dfa_lookahead): Likewise.
14650 (mips_avoid_hazard): Likewise.
14651 (mips_reorg_process_insns): Likewise.
14652 (mips_classify_branch_p6600): New function.
14653 * config/mips/mips.h (TUNE_P6600): New define.
14654 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
14655 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
14656 * config/mips/mips.md: Include p6600.md.
14657 (processor): Add p6600.
14658 * config/mips/p6600.md: New file.
14659 * doc/invoke.texi: Add p6600 to supported architectures.
14661 2018-06-13 Martin Sebor <msebor@redhat.com>
14663 PR tree-optimization/86114
14664 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
14666 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
14668 2018-06-13 Richard Biener <rguenther@suse.de>
14670 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
14671 Properly set vector type of the intermediate stmt.
14672 * tree-vect-stmts.c (vectorizable_operation): The destination
14673 var always has vectype_out type.
14675 2018-06-13 Jeff Law <law@redhat.com>
14677 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
14678 integer 0 for argument to print_rtl_with_bb.
14679 (rl78_reorg): Likewise.
14681 2018-06-13 David Malcolm <dmalcolm@redhat.com>
14683 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
14684 from rtx to rtx_insn *.
14685 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
14687 (add_sched_insns_for_speculation): Likewise for local "target",
14688 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
14689 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
14690 from rtx_insn ** to rtx_code_label **.
14691 (reorg_emit_nops): Likewise.
14692 (c6x_reorg): Likewise for local "call_labels".
14693 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
14695 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
14696 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
14697 the loops over LABEL_REFS.
14698 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
14700 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
14701 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
14702 (split_branches): Strengthen local "olabel" from rtx to
14703 rtx_insn *, adding a safe_as_a cast.
14704 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
14706 (add_insn_after): Likewise for first two params.
14707 (add_insn_before): Likewise.
14708 (remove_insn): Likewise for param.
14709 (emit_pattern_before_noloc): Likewise for second and third params.
14710 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
14711 (emit_call_insn_before_noloc): Likewise.
14712 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
14714 (emit_barrier_before): Likewise.
14715 (emit_label_before): Strengthen "label" param from "rtx" to
14716 "rtx_code_label *". Strengthen "before" param from "rtx" to
14718 (emit_insn_after_1): Strengthen "after" param from "rtx" to
14720 (emit_pattern_after_noloc): Likewise.
14721 (emit_insn_after_noloc): Likewise.
14722 (emit_jump_insn_after_noloc): Likewise.
14723 (emit_call_insn_after_noloc): Likewise.
14724 (emit_debug_insn_after_noloc): Likewise.
14725 (emit_barrier_after): Likewise.
14726 (emit_label_after): Likewise for both params.
14727 (emit_pattern_after_setloc): Likewise for "after" param. Convert
14728 "loc" param from "int" to "location_t".
14729 (emit_insn_after_setloc): Likewise.
14730 (emit_jump_insn_after_setloc): Likewise.
14731 (emit_call_insn_after_setloc): Likewise.
14732 (emit_debug_insn_after_setloc): Likewise.
14733 (emit_pattern_before_setloc): Likewise for "before" param. Convert
14734 "loc" param from "int" to "location_t".
14735 (emit_pattern_before): Convert NULL_RTX to NULL.
14736 (emit_insn_before_setloc): Convert "loc" param from "int" to
14738 (emit_jump_insn_before_setloc): Likewise.
14739 (emit_call_insn_before_setloc): Likewise.
14740 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
14741 rtx_insn *. Convert "loc" param from "int" to "location_t".
14742 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
14743 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
14744 Convert 3rd param from "int" to "location_t".
14745 (emit_barrier_before, emit_barrier_after, next_real_insn):
14746 Strengthen param from rtx to rtx_insn *.
14747 (emit_label_before): Strengthen 1st param from "rtx" to
14748 "rtx_code_label *". Strengthen 2nd param from "rtx" to
14750 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
14751 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
14752 Strengthen 2nd param from "rtx" to "rtx_insn *".
14753 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
14754 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
14755 Likewise. Convert 3rd param from "int" to "location_t".
14756 (emit_label_after): Strengthen 1st param from "rtx" to
14757 "rtx_code_label *".
14758 (next_real_insn, remove_insn): Strengthen param from "rtx" to
14760 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
14761 from "rtx" to "rtx_insn *".
14763 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
14765 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
14766 bodies streamed in with -Q.
14767 * dumpfile.c (dump_files): Add lto-stream-out dump file.
14768 * dumpfile.h (tree_dump_index): Add lto_stream_out.
14769 * gimple-streamer-out.c: Include gimple-pretty-print.h
14770 (output_bb): Dump stmts streamed.
14771 * lto-section-out.c: Include print-tree.h
14772 (lto_begin_section): Dump sections created.
14773 (lto_output_decl_index): Dump decl encoded.
14774 * lto-streamer-out.c: Include print-tree.h
14775 (create_output_block): Dump output block created.
14776 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
14777 (output_function): Dump function output.
14778 (output_constructor): Dump constructor streamed.
14779 (write_global_stream): Output indexes encoded.
14780 (produce_asm_for_decls): Dump streams encoded.
14781 * lto-streamer.c (streamer_dump_file): New global var.
14782 * lto-streamer.h (streamer_dump_file): Declare.
14783 * passes.c (ipa_write_summaries): Initialize streamer dump.
14784 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
14787 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
14790 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
14791 offsets for register save directives. Emit a second batch of save
14792 directives, if need be, when the function accesses prior frames.
14794 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
14796 * config/arc/fpu.md (fmasf4): Force operand to register.
14797 (fnmasf4): Likewise.
14799 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
14801 * config/arc/arc-protos.h (arc_pad_return): Remove.
14802 * config/arc/arc.c (machine_function): Remove force_short_suffix
14804 (arc_print_operand): Adjust printing of '&'.
14805 (arc_verify_short): Remove conditional printing of short suffix.
14806 (arc_final_prescan_insn): Remove reference to size_reason.
14807 (pad_return): New function.
14808 (arc_reorg): Call pad_return.
14809 (arc_pad_return): Remove.
14810 (arc_init_machine_status): Remove reference to force_short_suffix.
14811 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
14812 (attr length): When attribute iscompact is true force to 2
14813 regardless; in the case of maybe check if we want to force the
14814 instruction to have 4 bytes length.
14815 (nopv): Change it to generate 4 byte long nop as well.
14816 (blockage): New pattern.
14817 (simple_return): Remove call to arc_pad_return.
14818 (p_return_i): Likewise.
14820 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
14822 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
14824 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
14826 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
14829 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
14831 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
14832 for ARC700 and ARCv2.
14834 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
14837 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
14838 operands[2] instead of operands[1].
14841 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
14843 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
14844 case, check whether the outer register overlaps an unallocatable
14845 register, not just whether it fits the required class.
14847 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
14849 * poly-int.h (can_div_trunc_p): Add new overload in which all values
14851 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
14852 (memrefs_conflict_p): Likewise.
14853 (init_alias_analysis): Likewise.
14854 * cfgexpand.c (expand_debug_expr): Likewise.
14855 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
14856 * cse.c (fold_rtx): Likewise.
14857 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
14858 * expr.c (emit_block_move_hints): Likewise.
14859 (clear_storage_hints, push_block, emit_push_insn): Likewise.
14860 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
14861 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
14862 (emit_group_store): Likewise.
14863 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
14864 to read the PRE/POST_MODIFY increment.
14865 * calls.c (store_one_arg): Use strip_offset.
14866 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
14868 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
14870 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
14871 (simplify_binary_operation_1): Extend CONST_INT handling to
14873 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
14874 than a HOST_WIDE_INT.
14875 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
14877 (adjust_mems, add_stores): Update accodingly.
14878 (vt_canonicalize_addr): Track polynomial offsets.
14879 (emit_note_insn_var_location): Likewise.
14880 (vt_add_function_parameter): Likewise.
14881 (vt_initialize): Likewise.
14883 2018-06-12 Jeff Law <law@redhat.com>
14885 * config.gcc (alpha*-*-freebsd*): Remove.
14886 * config/alpha/freebsd.h: Remove.
14888 2018-06-12 David Malcolm <dmalcolm@redhat.com>
14891 * spellcheck-tree.c (levenshtein_distance): Rename to...
14892 (get_edit_distance): ...this, and update for underlying renaming.
14893 * spellcheck-tree.h (levenshtein_distance): Rename to...
14894 (get_edit_distance): ...this.
14895 * spellcheck.c (levenshtein_distance): Rename to...
14896 (get_edit_distance): ...this. Convert from Levenshtein distance
14897 to Damerau-Levenshtein distance by supporting transpositions of
14898 adjacent characters. Rename "v1" to "v_next" and "v0" to
14900 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
14901 (selftest::test_edit_distance_unit_test_oneway): ...this, and
14902 update for underlying renaming.
14903 (selftest::levenshtein_distance_unit_test): Rename to...
14904 (selftest::test_get_edit_distance_unit): ...this, and update for
14905 underlying renaming.
14906 (selftest::test_find_closest_string): Add example from PR 69968
14907 where transposition helps
14908 (selftest::test_metric_conditions): Update for renaming.
14909 (selftest::test_metric_conditions): Likewise.
14910 (selftest::spellcheck_c_tests): Likewise.
14911 * spellcheck.h (levenshtein_distance): Rename both overloads to...
14912 (get_edit_distance): ...this.
14913 (best_match::consider): Update for renaming.
14915 2018-06-12 Martin Sebor <msebor@redhat.com>
14917 PR tree-optimization/85259
14918 * builtins.c (compute_objsize): Handle constant offsets.
14919 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
14920 true iff a warning has been issued.
14921 * gimple.h (gimple_nonartificial_location): New function.
14922 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
14923 gimple_nonartificial_location and handle -Wno-system-headers.
14924 (handle_builtin_stxncpy): Same.
14926 2018-06-12 Martin Sebor <msebor@redhat.com>
14929 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
14931 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
14933 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
14934 BUILTIN_VEC_XST entries for pointer to double and long long.
14936 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
14939 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
14941 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
14944 2018-06-12 Martin Liska <mliska@suse.cz>
14946 * doc/options.texi: Document IntegerRange.
14948 2018-06-12 Martin Liska <mliska@suse.cz>
14950 * config/i386/i386.opt: Make MPX-related options as Deprecated.
14951 * opt-functions.awk: Handle Deprecated flag.
14952 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
14954 (read_cmdline_option): Report warning for a deprecated option.
14955 * opts.h (struct cl_option): Add new field cl_deprecated.
14956 (CL_ERR_DEPRECATED): New.
14958 2018-06-12 Martin Liska <mliska@suse.cz>
14960 * doc/options.texi: Document Deprecated option flag.
14962 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
14964 * config/arc/arc-arch.h (arc_extras): New enum.
14965 (arc_cpu_t):Add field extra.
14966 (arc_cpu_types): Consider the extras.
14967 * config/arc/arc-cpus.def: Add extras info.
14968 * config/arc/arc-opts.h (processor_type): Consider extra field.
14969 * config/arc/arc.c (arc_override_options): Handle extra field.
14971 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
14973 * config/arc/arc-arch.h: Update ARC_OPTX macro.
14974 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
14976 * config/arc/arc.c (arc_init): Update pic warning.
14977 (irq_range): Update irq range parsing warnings.
14978 (arc_override_options): Update various warning messages.
14979 (arc_handle_aux_attribute): Likewise.
14981 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
14983 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
14985 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
14987 * doc/sourcebuild.texi: Document usage of line number 0 in verify
14988 compiler messages directives.
14990 2018-06-12 Matthew Fortune <mfortune@gmail.com>
14992 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
14993 * config/mips/mips-tables.opt: Regenerate.
14994 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
14996 * doc/invoke.texi: Document -march=i6500.
14998 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
15000 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
15001 (i6400_gpmul): Add cpu_unit.
15002 (i6400_gpdiv): Likewise.
15003 (i6400_msa_add_d): Update reservations.
15004 (i6400_msa_int_add) Likewise.
15005 (i6400_msa_short_logic3) Likewise.
15006 (i6400_msa_short_logic2) Likewise.
15007 (i6400_msa_short_logic) Likewise.
15008 (i6400_msa_move) Likewise.
15009 (i6400_msa_cmp) Likewise.
15010 (i6400_msa_short_float2) Likewise.
15011 (i6400_msa_div_d) Likewise.
15012 (i6400_msa_long_logic1) Likewise.
15013 (i6400_msa_long_logic2) Likewise.
15014 (i6400_msa_mult) Likewise.
15015 (i6400_msa_long_float2) Likewise.
15016 (i6400_msa_long_float4) Likewise.
15017 (i6400_msa_long_float5) Likewise.
15018 (i6400_msa_long_float8) Likewise.
15019 (i6400_fpu_fadd): Include frint type.
15020 (i6400_fpu_store): New define_insn_reservation.
15021 (i6400_fpu_load): Likewise.
15022 (i6400_fpu_move): Likewise.
15023 (i6400_fpu_fcmp): Likewise.
15024 (i6400_fpu_fmadd): Likewise.
15025 (i6400_int_mult): Include imul3nc type and update reservation.
15026 (i6400_int_div): Include idiv3 type and update reservation.
15027 (i6400_int_load): Update to check type not move_type.
15028 (i6400_int_store): Likewise.
15029 (i6400_int_prefetch): Set zero latency.
15031 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
15033 * gcc.c: Document new %@{...} sequence.
15034 (LINK_COMMAND_SPEC): Use it for the -L switches.
15035 (cpp_unique_options): Use it for the -I switches.
15036 (at_file_argbuf): New global variable.
15037 (in_at_file): Likewise.
15038 (alloc_args): Create at_file_argbuf.
15039 (clear_args): Truncate at_file_argbuf.
15040 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
15041 (open_at_file): New function.
15042 (close_at_file): Likewise.
15043 (create_at_file): Delete.
15044 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
15047 (validate_switches_from_spec): Deal with %@{...} sequence.
15048 (validate_switches): Likewise.
15049 (driver::finalize): Call clear_args.
15051 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
15053 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
15055 2018-06-11 Martin Sebor <msebor@redhat.com>
15057 * doc/invoke.texi (-Wall): List -Wc++17-compat.
15058 (Wno-class-memaccess): Add @opindex.
15059 (Wno-templates, Wno-multiple-inheritance): Same.
15060 (Wno-virtual-inheritance, Wno-namespaces): Same.
15061 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
15062 (Wno-format-overflow, Wno-format-truncation): Same.
15063 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
15064 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
15065 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
15066 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
15067 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
15068 (Wno-misspelled-isr): Same.
15070 2018-06-11 Martin Sebor <msebor@redhat.com>
15072 * PR tree-optimization/86083
15073 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
15075 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
15077 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
15079 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
15082 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
15083 on the correct operand.
15084 (*movdi_internal64): Ditto.
15086 2018-06-11 Martin Liska <mliska@suse.cz>
15088 PR tree-optimization/86089
15089 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
15091 2018-06-11 Julia Koval <julia.koval@intel.com>
15093 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
15094 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
15095 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
15097 2018-06-11 Olivier Hainque <hainque@adacore.com>
15099 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
15100 for Ada with strict dwarf2.
15102 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
15105 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
15108 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
15110 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
15112 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
15114 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
15117 2018-06-08 Martin Liska <mliska@suse.cz>
15119 * tree-cfg.h (debug_function): Fix argument type to match
15122 2018-06-08 Martin Liska <mliska@suse.cz>
15124 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
15125 Remove usage of MPX-related (and removed) fields.
15126 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
15128 2018-06-08 David Malcolm <dmalcolm@redhat.com>
15130 * cfg.c (debug): Use TDF_NONE rather than 0.
15131 * cfghooks.c (debug): Likewise.
15132 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
15133 (struct dump_option_value_info): Convert to...
15134 (struct kv_pair): ...this template type.
15135 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
15137 (optinfo_verbosity_options): Likewise.
15138 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
15140 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
15141 than int for "optgroup_flags" param.
15142 (dump_generic_expr_loc): Use dump_flags_t rather than int for
15144 (dump_dec): Likewise.
15145 (dump_finish): Use TDF_NONE rather than 0.
15146 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
15147 rather than int for "optgroup_flags" param. Use TDF_NONE rather
15148 than 0. Update for change to option_ptr.
15149 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
15150 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
15151 0. Update for changes to optinfo_verbosity_options and
15153 (opt_info_switch_p): Convert optgroup_flags from int to
15155 (dump_basic_block): Use dump_flags_t rather than int
15156 for "dump_kind" param.
15157 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
15158 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
15159 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
15160 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
15161 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
15162 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
15163 TDF_NONE): Convert from macros to...
15164 (enum dump_flag): ...this new enum.
15165 (dump_flags_t): Update to use enum.
15166 (operator|, operator&, operator~, operator|=, operator&=):
15167 Implement for dump_flags_t.
15168 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
15169 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
15170 Convert from macros to...
15171 (enum optgroup_flag): ...this new enum.
15172 (optgroup_flags_t): New typedef.
15173 (operator|, operator|=): Implement for optgroup_flags_t.
15174 (struct dump_file_info): Convert field "alt_flags" to
15175 dump_flags_t. Convert field "optgroup_flags" to
15177 (dump_basic_block): Use dump_flags_t rather than int for param.
15178 (dump_generic_expr_loc): Likewise.
15179 (dump_dec): Likewise.
15180 (dump_register): Convert param "optgroup_flags" to
15182 (opt_info_enable_passes): Likewise.
15183 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
15185 * gimple-pretty-print.c (debug): Likewise.
15186 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
15187 (merged_store_group::apply_stores): Likewise.
15188 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
15189 * gimple.c (verify_gimple_pp): Likewise.
15190 * graphite-poly.c (print_pbb_body): Likewise.
15191 * passes.c (pass_manager::register_one_dump_file): Convert
15192 local "optgroup_flags" to optgroup_flags_t.
15193 * print-tree.c (print_node): Use TDF_NONE rather than 0.
15195 (debug_body): Likewise.
15196 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
15197 to optgroup_flags_t.
15198 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
15200 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
15201 (convert_mult_to_fma): Likewise.
15202 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
15203 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
15204 * tree-vect-data-refs.c (dump_lower_bound): Convert param
15205 "dump_kind" to dump_flags_t.
15207 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
15209 * config/rs6000/rs6000.c (min, max): Delete.
15211 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
15213 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
15214 -mabi=spe and -mabi=no-spe.
15216 2018-06-08 Martin Liska <mliska@suse.cz>
15218 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
15219 where we expect an existing summary.
15221 2018-06-08 Martin Liska <mliska@suse.cz>
15223 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
15224 * ipa-inline.h (estimate_edge_growth): Likewise.
15226 2018-06-08 Martin Liska <mliska@suse.cz>
15228 * cgraph.c (function_version_hasher::hash): Use
15229 cgraph_node::get_uid ().
15230 (function_version_hasher::equal):
15231 * cgraph.h (cgraph_node::get_uid): New method.
15232 * ipa-inline.c (update_caller_keys): Use
15233 cgraph_node::get_uid ().
15234 (update_callee_keys): Likewise.
15235 * ipa-utils.c (searchc): Likewise.
15236 (ipa_reduced_postorder): Likewise.
15237 * lto-cgraph.c (input_node): Likewise.
15238 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
15239 * symbol-summary.h (symtab_insertion): Likewise.
15240 (symtab_removal): Likewise.
15241 (symtab_duplication): Likewise.
15242 * tree-pretty-print.c (dump_function_header): Likewise.
15243 * tree-sra.c (convert_callers_for_node): Likewise.
15245 2018-06-08 Martin Liska <mliska@suse.cz>
15247 * cgraph.c (symbol_table::create_edge): Always assign a new
15249 (symbol_table::free_edge): Do not recycle numbers.
15250 * cgraph.h (cgraph_edge::get): New method.
15251 * symbol-summary.h (symtab_removal): Use it.
15252 (symtab_duplication): Likewise.
15253 (call_summary::hashable_uid): Remove.
15255 2018-06-08 Martin Liska <mliska@suse.cz>
15257 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
15258 (initialize_growth_caches): Remove.
15259 (free_growth_caches): Likewise.
15260 (do_estimate_edge_time): Use edge_growth_cache.
15261 (do_estimate_edge_size): Likewise.
15262 (do_estimate_edge_hints): Likewise.
15263 * ipa-inline.c (reset_edge_caches): Likewise.
15264 (recursive_inlining): Likewise.
15265 (inline_small_functions): Likewise.
15266 * ipa-inline.h (initialize_growth_caches): Remove.
15267 (estimate_edge_size): Likewise.
15268 (estimate_edge_time): Likewise.
15269 (estimate_edge_hints): Likewise.
15270 (reset_edge_growth_cache): Likewise.
15271 * symbol-summary.h (call_summary::remove): New method.
15273 2018-06-08 Martin Liska <mliska@suse.cz>
15275 * ipa-cp.c (class edge_clone_summary): New summary.
15276 (grow_edge_clone_vectors): Remove.
15277 (ipcp_edge_duplication_hook): Remove.
15278 (class edge_clone_summary_t): New call_summary class.
15279 (ipcp_edge_removal_hook): Remove.
15280 (edge_clone_summary_t::duplicate): New function.
15281 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
15282 (create_specialized_node): Likewise.
15283 (ipcp_driver): Initialize edge_clone_summaries and do not
15286 2018-06-08 Martin Liska <mliska@suse.cz>
15288 * symbol-summary.h (get): New function.
15289 (call_summary::m_initialize_when_cloning): New class member.
15291 2018-06-08 Martin Liska <mliska@suse.cz>
15293 * cgraph.c (cgraph_node::remove): Do not recycle uid.
15294 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
15295 (symbol_table::allocate_cgraph_symbol): Do not set uid.
15296 * passes.c (uid_hash_t): Record removed_nodes by their uids.
15297 (remove_cgraph_node_from_order): Use the removed_nodes set.
15298 (do_per_function_toporder): Likwise.
15299 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
15300 instead of summary_uid.
15301 (symtab_removal): Likewise.
15302 (symtab_duplication): Likewise.
15304 2018-06-08 Martin Liska <mliska@suse.cz>
15306 * ipa-cp.c (ipcp_store_bits_results): Use
15307 ipcp_transformation_sum.
15308 (ipcp_store_vr_results): Likewise.
15309 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
15311 (ipcp_transformation_initialize): ... this.
15312 (ipa_set_node_agg_value_chain):
15313 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
15314 (write_ipcp_transformation_info): Likewise.
15315 (read_ipcp_transformation_info): Likewise.
15316 (ipcp_update_bits): Likewise.
15317 (ipcp_update_vr): Likewise.
15318 (ipcp_transform_function): Likewise.
15319 * ipa-prop.h: Rename ipcp_transformation_summary to
15320 ipcp_transformation.
15321 (class ipcp_transformation_t): New function summary.
15322 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
15323 (ipa_get_agg_replacements_for_node): Likewise.
15325 2018-06-08 Martin Liska <mliska@suse.cz>
15327 * ipa-pure-const.c (struct funct_state_d): Do it class instead
15329 (class funct_state_summary_t): New function_summary class.
15330 (has_function_state): Remove.
15331 (get_function_state): Likewise.
15332 (set_function_state): Likewise.
15333 (add_new_function): Likewise.
15334 (funct_state_summary_t::insert): New function.
15335 (duplicate_node_data): Remove.
15336 (remove_node_data): Remove.
15337 (funct_state_summary_t::duplicate): New function.
15338 (register_hooks): Create new funct_state_summaries.
15339 (pure_const_generate_summary): Use it.
15340 (pure_const_write_summary): Likewise.
15341 (pure_const_read_summary): Likewise.
15342 (propagate_pure_const): Likewise.
15343 (propagate_nothrow): Likewise.
15344 (dump_malloc_lattice): Likewise.
15345 (propagate_malloc): Likewise.
15346 (execute): Do not register hooks, just remove summary
15348 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
15351 2018-06-08 Martin Liska <mliska@suse.cz>
15353 * ipa-reference.c (remove_node_data): Remove.
15354 (duplicate_node_data): Likewise.
15355 (class ipa_ref_var_info_summary_t): New class.
15356 (class ipa_ref_opt_summary_t): Likewise.
15357 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
15358 (get_reference_optimization_summary): Use
15359 ipa_ref_opt_sum_summaries.
15360 (set_reference_vars_info): Remove.
15361 (set_reference_optimization_summary): Likewise.
15362 (ipa_init): Create summaries.
15363 (init_function_info): Use function summary.
15364 (ipa_ref_opt_summary_t::duplicate): New function.
15365 (ipa_ref_opt_summary_t::remove): New function.
15366 (get_read_write_all_from_node): Fix GNU coding style.
15367 (propagate): Use function summary.
15368 (write_node_summary_p): Fix GNU coding style.
15369 (stream_out_bitmap): Likewise.
15370 (ipa_reference_read_optimization_summary): Use function summary.
15371 (ipa_reference_c_finalize): Do not release hooks.
15373 2018-06-08 Martin Liska <mliska@suse.cz>
15375 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
15376 (analyze_function_body): Extract multiple calls of get_create.
15377 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
15378 * ipa-inline.c (recursive_inlining): Use ::get method.
15379 * ipa-inline.h (estimate_edge_growth): Likewise.
15381 2018-06-08 Martin Liska <mliska@suse.cz>
15383 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
15385 (hsa_function_summary::hsa_function_summary): Use the new enum
15387 (hsa_gpu_implementation_p): Use hsa_summaries::get.
15388 * hsa-gen.c (hsa_get_host_function): Likewise.
15389 (get_brig_function_name): Likewise.
15390 * ipa-hsa.c (process_hsa_functions): Likewise.
15391 (ipa_hsa_write_summary): Likewise.
15392 * symbol-summary.h (symtab_duplication): Use ::get function/
15393 (get): New function.
15395 2018-06-08 Martin Liska <mliska@suse.cz>
15397 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
15399 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
15400 (hsa_register_kernel): Likewise.
15401 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
15402 * hsa-gen.c (hsa_get_host_function): Likewise.
15403 (get_brig_function_name): Likewise.
15404 (generate_hsa): Likewise.
15405 (pass_gen_hsail::execute): Likewise.
15406 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
15407 (devirtualization_time_bonus): Likewise.
15408 (ipcp_propagate_stage): Likewise.
15409 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
15410 (edge_set_predicate): Likewise.
15411 (evaluate_conditions_for_known_args): Likewise.
15412 (evaluate_properties_for_edge): Likewise.
15413 (ipa_fn_summary::reset): Likewise.
15414 (ipa_fn_summary_t::duplicate): Likewise.
15415 (dump_ipa_call_summary): Likewise.
15416 (ipa_dump_fn_summary): Likewise.
15417 (analyze_function_body): Likewise.
15418 (compute_fn_summary): Likewise.
15419 (estimate_edge_devirt_benefit): Likewise.
15420 (estimate_edge_size_and_time): Likewise.
15421 (estimate_calls_size_and_time): Likewise.
15422 (estimate_node_size_and_time): Likewise.
15423 (inline_update_callee_summaries): Likewise.
15424 (remap_edge_change_prob): Likewise.
15425 (remap_edge_summaries): Likewise.
15426 (ipa_merge_fn_summary_after_inlining): Likewise.
15427 (ipa_update_overall_fn_summary): Likewise.
15428 (read_ipa_call_summary): Likewise.
15429 (inline_read_section): Likewise.
15430 (write_ipa_call_summary): Likewise.
15431 (ipa_fn_summary_write): Likewise.
15432 (ipa_free_fn_summary): Likewise.
15433 * ipa-hsa.c (process_hsa_functions): Likewise.
15434 (ipa_hsa_write_summary): Likewise.
15435 (ipa_hsa_read_section): Likewise.
15436 * ipa-icf.c (sem_function::merge): Likewise.
15437 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
15438 (do_estimate_edge_time): Likewise.
15439 (estimate_size_after_inlining): Likewise.
15440 (estimate_growth): Likewise.
15441 (growth_likely_positive): Likewise.
15442 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
15443 (inline_call): Likewise.
15444 * ipa-inline.c (caller_growth_limits): Likewise.
15445 (can_inline_edge_p): Likewise.
15446 (can_inline_edge_by_limits_p): Likewise.
15447 (compute_uninlined_call_time): Likewise.
15448 (compute_inlined_call_time): Likewise.
15449 (want_inline_small_function_p): Likewise.
15450 (edge_badness): Likewise.
15451 (update_caller_keys): Likewise.
15452 (update_callee_keys): Likewise.
15453 (recursive_inlining): Likewise.
15454 (inline_small_functions): Likewise.
15455 (inline_to_all_callers_1): Likewise.
15456 (dump_overall_stats): Likewise.
15457 (early_inline_small_functions): Likewise.
15458 (early_inliner): Likewise.
15459 * ipa-inline.h (estimate_edge_growth): Likewise.
15460 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
15461 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15462 * ipa-prop.h (IPA_NODE_REF): Likewise.
15463 (IPA_EDGE_REF): Likewise.
15464 * ipa-pure-const.c (malloc_candidate_p): Likewise.
15465 (propagate_malloc): Likewise.
15466 * ipa-split.c (execute_split_functions): Likewise.
15467 * symbol-summary.h: Rename get to get_create.
15469 (get_create): Likewise.
15470 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
15472 2018-06-08 Martin Liska <mliska@suse.cz>
15474 * symbol-summary.h (release): Move definition out of class
15476 (symtab_removal): Likewise.
15477 (symtab_duplication): Likewise.
15479 2018-06-08 Martin Liska <mliska@suse.cz>
15481 * symbol-summary.h (function_summary): Move constructor
15482 implementation out of class declaration.
15483 (release): Likewise.
15484 (symtab_insertion): Likewise.
15485 (symtab_removal): Likewise.
15486 (symtab_duplication): Likewise.
15489 2018-06-08 Martin Liska <mliska@suse.cz>
15491 * Makefile.in: Remove support for MPX (macros, related functions,
15492 fields in cgraph_node, ...).
15493 * builtin-types.def (BT_BND): Likewise.
15494 (BT_FN_BND_CONST_PTR): Likewise.
15495 (BT_FN_CONST_PTR_BND): Likewise.
15496 (BT_FN_VOID_PTR_BND): Likewise.
15497 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
15498 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
15499 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
15500 (expand_builtin_mempcpy_with_bounds): Likewise.
15501 (expand_builtin_memset_with_bounds): Likewise.
15502 (expand_builtin_memset_args): Likewise.
15503 (std_expand_builtin_va_start): Likewise.
15504 (expand_builtin): Likewise.
15505 (expand_builtin_with_bounds): Likewise.
15506 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
15507 (DEF_LIB_BUILTIN_CHKP): Likewise.
15508 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
15509 (DEF_CHKP_BUILTIN): Likewise.
15510 (BUILT_IN_MEMCPY): Likewise.
15511 (BUILT_IN_MEMMOVE): Likewise.
15512 (BUILT_IN_MEMPCPY): Likewise.
15513 (BUILT_IN_MEMSET): Likewise.
15514 (BUILT_IN_STPCPY): Likewise.
15515 (BUILT_IN_STRCAT): Likewise.
15516 (BUILT_IN_STRCHR): Likewise.
15517 (BUILT_IN_STRCPY): Likewise.
15518 (BUILT_IN_STRLEN): Likewise.
15519 (BUILT_IN_MEMCPY_CHK): Likewise.
15520 (BUILT_IN_MEMMOVE_CHK): Likewise.
15521 (BUILT_IN_MEMPCPY_CHK): Likewise.
15522 (BUILT_IN_MEMSET_CHK): Likewise.
15523 (BUILT_IN_STPCPY_CHK): Likewise.
15524 (BUILT_IN_STRCAT_CHK): Likewise.
15525 (BUILT_IN_STRCPY_CHK): Likewise.
15526 * calls.c (store_bounds): Likewise.
15527 (emit_call_1): Likewise.
15528 (special_function_p): Likewise.
15529 (maybe_warn_nonstring_arg): Likewise.
15530 (initialize_argument_information): Likewise.
15531 (finalize_must_preallocate): Likewise.
15532 (compute_argument_addresses): Likewise.
15533 (expand_call): Likewise.
15534 * cfgexpand.c (expand_call_stmt): Likewise.
15535 (expand_return): Likewise.
15536 (expand_gimple_stmt_1): Likewise.
15537 (pass_expand::execute): Likewise.
15538 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
15539 (cgraph_node::remove): Likewise.
15540 (cgraph_node::dump): Likewise.
15541 (cgraph_node::verify_node): Likewise.
15542 * cgraph.h (chkp_function_instrumented_p): Likewise.
15543 (symtab_node::get_alias_target): Likewise.
15544 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
15545 (cgraph_local_p): Likewise.
15546 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
15547 (cgraph_edge::rebuild_references): Likewise.
15548 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
15549 (walk_polymorphic_call_targets): Likewise.
15550 (cgraph_node::expand_thunk): Likewise.
15551 (symbol_table::output_weakrefs): Likewise.
15552 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
15553 (ix86_handle_option): Likewise.
15554 * config/i386/constraints.md: Likewise.
15555 * config/i386/i386-builtin-types.def (BND): Likewise.
15559 * config/i386/i386-builtin.def (BDESC_END): Likewise.
15560 (BDESC_FIRST): Likewise.
15562 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
15563 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
15564 * config/i386/i386.c (enum reg_class): Likewise.
15565 (ix86_target_string): Likewise.
15566 (ix86_option_override_internal): Likewise.
15567 (ix86_conditional_register_usage): Likewise.
15568 (ix86_valid_target_attribute_inner_p): Likewise.
15569 (ix86_set_indirect_branch_type): Likewise.
15570 (ix86_set_current_function): Likewise.
15571 (ix86_function_arg_regno_p): Likewise.
15572 (init_cumulative_args): Likewise.
15573 (ix86_function_arg_advance): Likewise.
15574 (ix86_function_arg): Likewise.
15575 (ix86_pass_by_reference): Likewise.
15576 (ix86_function_value_regno_p): Likewise.
15577 (ix86_function_value_1): Likewise.
15578 (ix86_function_value_bounds): Likewise.
15579 (ix86_return_in_memory): Likewise.
15580 (ix86_setup_incoming_vararg_bounds): Likewise.
15581 (ix86_va_start): Likewise.
15582 (indirect_thunk_need_prefix): Likewise.
15583 (print_reg): Likewise.
15584 (ix86_print_operand): Likewise.
15585 (ix86_expand_call): Likewise.
15586 (ix86_output_function_return): Likewise.
15587 (reg_encoded_number): Likewise.
15588 (BDESC_VERIFYS): Likewise.
15589 (ix86_init_mpx_builtins): Likewise.
15590 (ix86_init_builtins): Likewise.
15591 (ix86_emit_cmove): Likewise.
15592 (ix86_emit_move_max): Likewise.
15593 (ix86_expand_builtin): Likewise.
15594 (ix86_builtin_mpx_function): Likewise.
15595 (ix86_get_arg_address_for_bt): Likewise.
15596 (ix86_load_bounds): Likewise.
15597 (ix86_store_bounds): Likewise.
15598 (ix86_load_returned_bounds): Likewise.
15599 (ix86_store_returned_bounds): Likewise.
15600 (ix86_class_likely_spilled_p): Likewise.
15601 (ix86_hard_regno_mode_ok): Likewise.
15602 (x86_order_regs_for_local_alloc): Likewise.
15603 (ix86_mitigate_rop): Likewise.
15604 (ix86_bnd_prefixed_insn_p): Likewise.
15605 (ix86_mpx_bound_mode): Likewise.
15606 (ix86_make_bounds_constant): Likewise.
15607 (ix86_initialize_bounds): Likewise.
15608 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
15609 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
15610 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
15611 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
15612 (TARGET_CHKP_BOUND_MODE): Likewise.
15613 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
15614 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
15615 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
15616 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
15617 * config/i386/i386.h (TARGET_MPX): Likewise.
15618 (TARGET_MPX_P): Likewise.
15619 (VALID_BND_REG_MODE): Likewise.
15620 (FIRST_BND_REG): Likewise.
15621 (LAST_BND_REG): Likewise.
15622 (enum reg_class): Likewise.
15623 (BND_REG_P): Likewise.
15624 (BND_REGNO_P): Likewise.
15625 (BNDmode): Likewise.
15626 (ADJUST_INSN_LENGTH): Likewise.
15627 * config/i386/i386.md: Likewise.
15628 * config/i386/i386.opt: Likewise.
15629 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
15630 (defined): Likewise.
15631 (LINK_MPX): Likewise.
15632 (MPX_SPEC): Likewise.
15633 (LIBMPX_SPEC): Likewise.
15634 (LIBMPXWRAPPERS_SPEC): Likewise.
15635 (CHKP_SPEC): Likewise.
15636 * config/i386/predicates.md: Likewise.
15637 * dbxout.c (dbxout_type): Likewise.
15638 * doc/extend.texi: Likewise.
15639 * doc/invoke.texi: Likewise.
15640 * doc/md.texi: Likewise.
15641 * doc/tm.texi: Likewise.
15642 * doc/tm.texi.in: Likewise.
15643 * dwarf2out.c (is_base_type): Likewise.
15644 (gen_formal_types_die): Likewise.
15645 (gen_subprogram_die): Likewise.
15646 (gen_type_die_with_usage): Likewise.
15647 (gen_decl_die): Likewise.
15648 (dwarf2out_late_global_decl): Likewise.
15649 * expr.c (expand_assignment): Likewise.
15650 (emit_storent_insn): Likewise.
15651 (store_expr_with_bounds): Likewise.
15652 (store_expr): Likewise.
15653 (expand_expr_real_1): Likewise.
15654 * expr.h (store_expr_with_bounds): Likewise.
15655 * function.c (use_register_for_decl): Likewise.
15656 (struct bounds_parm_data): Likewise.
15657 (assign_parms_augmented_arg_list): Likewise.
15658 (assign_parm_find_entry_rtl): Likewise.
15659 (assign_parm_is_stack_parm): Likewise.
15660 (assign_parm_load_bounds): Likewise.
15661 (assign_bounds): Likewise.
15662 (assign_parms): Likewise.
15663 (expand_function_start): Likewise.
15664 * gcc.c (CHKP_SPEC): Likewise.
15665 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
15666 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
15667 (wrestrict_dom_walker::check_call): Likewise.
15668 * gimple.c (gimple_build_call_from_tree): Likewise.
15669 * gimple.h (enum gf_mask): Likewise.
15670 (gimple_call_with_bounds_p): Likewise.
15671 (gimple_call_set_with_bounds): Likewise.
15672 * gimplify.c (gimplify_init_constructor): Likewise.
15673 * ipa-cp.c (initialize_node_lattices): Likewise.
15674 (propagate_constants_across_call): Likewise.
15675 (find_more_scalar_values_for_callers_subset): Likewise.
15676 * ipa-hsa.c (process_hsa_functions): Likewise.
15677 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
15678 * ipa-icf.c (sem_function::merge): Likewise.
15679 * ipa-inline.c (early_inliner): Likewise.
15680 * ipa-pure-const.c (warn_function_noreturn): Likewise.
15681 (warn_function_cold): Likewise.
15682 (propagate_pure_const): Likewise.
15683 * ipa-ref.h (enum GTY): Likewise.
15684 * ipa-split.c (find_retbnd): Likewise.
15685 (consider_split): Likewise.
15686 (split_function): Likewise.
15687 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
15688 * ipa.c (walk_polymorphic_call_targets): Likewise.
15689 (symbol_table::remove_unreachable_nodes): Likewise.
15690 (process_references): Likewise.
15691 (cgraph_build_static_cdtor_1): Likewise.
15692 * lto-cgraph.c (lto_output_node): Likewise.
15693 (output_refs): Likewise.
15694 (compute_ltrans_boundary): Likewise.
15695 (input_overwrite_node): Likewise.
15696 (input_node): Likewise.
15697 (input_cgraph_1): Likewise.
15698 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
15699 * passes.c (pass_manager::execute_early_local_passes): Likewise.
15700 (class pass_chkp_instrumentation_passes): Likewise.
15701 (make_pass_chkp_instrumentation_passes): Likewise.
15702 * passes.def: Likewise.
15703 * rtl.h (struct GTY): Likewise.
15704 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
15705 * stor-layout.c (layout_type): Likewise.
15706 * symtab.c: Likewise.
15707 * target.def: Likewise.
15708 * targhooks.c (default_chkp_bound_type): Likewise.
15709 (default_chkp_bound_mode): Likewise.
15710 (default_builtin_chkp_function): Likewise.
15711 (default_chkp_function_value_bounds): Likewise.
15712 (default_chkp_make_bounds_constant): Likewise.
15713 (default_chkp_initialize_bounds): Likewise.
15714 * targhooks.h (default_chkp_bound_type): Likewise.
15715 (default_chkp_bound_mode): Likewise.
15716 (default_builtin_chkp_function): Likewise.
15717 (default_chkp_function_value_bounds): Likewise.
15718 (default_chkp_make_bounds_constant): Likewise.
15719 (default_chkp_initialize_bounds): Likewise.
15720 * toplev.c (compile_file): Likewise.
15721 (process_options): Likewise.
15722 * tree-core.h (DEF_BUILTIN): Likewise.
15723 (DEF_BUILTIN_CHKP): Likewise.
15724 * tree-inline.c (declare_return_variable): Likewise.
15725 (remap_gimple_stmt): Likewise.
15726 (copy_bb): Likewise.
15727 (initialize_inlined_parameters): Likewise.
15728 (expand_call_inline): Likewise.
15729 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
15730 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
15731 (make_pass_ipa_chkp_produce_thunks): Likewise.
15732 (make_pass_chkp): Likewise.
15733 (make_pass_chkp_opt): Likewise.
15734 (make_pass_chkp_instrumentation_passes): Likewise.
15735 * tree-pretty-print.c (dump_generic_node): Likewise.
15736 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
15737 * tree-ssa-dce.c (propagate_necessity): Likewise.
15738 (eliminate_unnecessary_stmts): Likewise.
15739 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
15740 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
15741 * tree-ssa-sccvn.h: Likewise.
15742 * tree-ssa-strlen.c (get_string_length): Likewise.
15743 (valid_builtin_call): Likewise.
15744 (adjust_last_stmt): Likewise.
15745 (handle_builtin_strchr): Likewise.
15746 (handle_builtin_strcpy): Likewise.
15747 (handle_builtin_stxncpy): Likewise.
15748 (handle_builtin_memcpy): Likewise.
15749 (handle_builtin_strcat): Likewise.
15750 (strlen_check_and_optimize_stmt): Likewise.
15751 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
15752 * tree-streamer-in.c: Likewise.
15753 * tree-streamer.c (record_common_node): Likewise.
15754 * tree.c (tree_code_size): Likewise.
15755 (wide_int_to_tree_1): Likewise.
15756 (type_contains_placeholder_1): Likewise.
15757 (build_common_tree_nodes): Likewise.
15758 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
15759 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
15760 (POINTER_BOUNDS_P): Likewise.
15761 (BOUNDED_TYPE_P): Likewise.
15762 (BOUNDED_P): Likewise.
15763 (CALL_WITH_BOUNDS_P): Likewise.
15764 (pointer_bounds_type_node): Likewise.
15765 * value-prof.c (gimple_ic): Likewise.
15766 * var-tracking.c (vt_add_function_parameters): Likewise.
15767 * varasm.c (make_decl_rtl): Likewise.
15768 (assemble_start_function): Likewise.
15769 (output_constant): Likewise.
15770 (maybe_assemble_visibility): Likewise.
15771 * varpool.c (ctor_for_folding): Likewise.
15772 * chkp-builtins.def: Remove.
15773 * ipa-chkp.c: Remove.
15774 * ipa-chkp.h: Remove.
15775 * rtl-chkp.c: Remove.
15776 * rtl-chkp.h: Remove.
15777 * tree-chkp-opt.c: Remove.
15778 * tree-chkp.c: Remove.
15779 * tree-chkp.h: Remove.
15781 2018-06-07 Carl Love <cel@us.ibm.com>
15783 * config/rs6000/vsx.md (vextract_fp_from_shorth,
15784 vextract_fp_from_shortl): Add BE support.
15786 2018-06-07 Paul Koning <ni1d@arrl.net>
15788 * compare-elim.c (try_merge_compare): Don't merge compare if
15789 address contains a side effect.
15790 (try_eliminate_compare): Likewise.
15792 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
15794 * config.gcc: Support "tremont".
15795 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
15796 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15798 * config/i386/i386.c (m_TREMONT): Define.
15799 (processor_target_table): Add "tremont".
15800 (PTA_TREMONT): Define.
15801 (ix86_lea_outperforms): Add TARGET_TREMONT.
15802 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
15803 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
15804 and M_INTEL_GOLDMONT_PLUS.
15805 (fold_builtin_cpu): Add "tremont".
15806 (ix86_add_stmt_cost): Add TARGET_TREMONT.
15807 (ix86_option_override_internal): Add "tremont".
15808 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
15809 (processor_type): Add PROCESSOR_TREMONT.
15810 * config/i386/x86-tune.def: Add m_TREMONT.
15811 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
15813 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
15815 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
15816 symbol defined for msp430i* devices to be lower case.
15818 2018-06-07 Richard Biener <rguenther@suse.de>
15820 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
15821 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
15822 Properly wrap signed arithmetic if overflow wraps.
15824 2018-06-07 Jakub Jelinek <jakub@redhat.com>
15826 PR tree-optimization/69615
15827 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
15828 of a cast from a same precision integral SSA_NAME in a bb dominated
15829 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
15830 cast to utype if rhs2 has already a compatible type.
15832 2018-06-07 Richard Biener <rguenther@suse.de>
15834 PR tree-optimization/85935
15835 * graphite-scop-detection.c (find_params_in_bb): Analyze
15836 condition operands with respect to the correct loop. Assert
15837 the analysis doesn't fail.
15839 2018-06-04 Carl Love <cel@us.ibm.com>
15841 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
15842 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
15843 as it is slightly cheaper.
15844 (first_match_or_eos_index_<mode>):
15845 Calculate index using natural element order.
15846 (first_match_index_<mode>):
15847 Calculate index using natural element order.
15848 (first_match_or_eos_index_<mode>):
15849 Calculate index using natural order.
15850 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
15851 for BE and LE modes.
15852 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
15853 P9V_BUILTIN_VCLZLSBB_V16QI.
15854 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
15857 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
15859 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
15860 indentation and line wrap for many prototypes. Add missing
15861 @smallexample directives around block of prototypes for vec_xl and
15864 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
15866 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
15867 track if we pass or return IEEE 128-bit floating point.
15868 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
15869 C++ mangling that is compatible with GCC 8.1.
15870 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
15871 (init_cumulative_args): Note if we pass or return IEEE 128-bit
15872 floating point types.
15873 (rs6000_function_arg_advance_1): Likewise.
15874 (rs6000_mangle_type): Optionally generate mangled names that match
15875 what GCC 8.1 generated for IEEE 128-bit floating point types.
15876 (rs6000_globalize_decl_name): If we have an external function that
15877 passes or returns IEEE 128-bit types, generate a weak reference
15878 from the mangled name used in GCC 8.1 to the current mangled
15880 (rs6000_init_builtins): Make __ibm128 use the long double type if
15881 long double is IBM extended double. Make __float128 use the long
15882 double type if long double is IEEE 128-bit.
15885 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
15886 macro for __ibm128 built-in functions.
15887 (PACK_IF): Add __ibm128 pack/unpack functions.
15888 (UNPACK_IF): Likewise.
15889 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
15890 enable long double built-in functions if long double is IEEE
15891 128-bit floating point.
15892 (rs6000_invalid_builtin): Update long double built-in function
15894 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
15895 functions, adjust the built-in function to use the long double
15896 built-in function if __ibm128 and long double are the same type.
15897 * doc/extend.texi (PowerPC builtins): Update documention for
15898 __builtin_{,un}pack_longdouble. Add documentation for
15899 __builtin_{,un}pack_ibm128.
15901 2018-06-06 Jim Wilson <jimw@sifive.com>
15903 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
15904 (struct machine_function): New field interrupt_mode.
15905 (riscv_handle_type_attribute): New function. Add forward declaration.
15906 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
15907 (riscv_expand_epilogue): Check interrupt_mode field.
15908 (riscv_set_current_function): Check interrupt attribute args and
15909 set interrupt_mode field.
15910 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
15911 (riscv_sret, riscv_uret): New.
15912 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
15913 new arguments to interrupt attribute.
15915 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
15918 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
15919 Don't handle -mcpu=power8 if -mpower9-vector is also used.
15921 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
15923 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
15924 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
15925 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
15926 several redundant entries.
15928 2018-06-06 David Malcolm <dmalcolm@redhat.com>
15930 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
15931 type from "rtx" to "rtx_insn *".
15932 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
15933 for local "call_insn", removing cast.
15934 (ix86_expand_call): Likewise, introducing a "call_insn" local.
15936 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
15938 PR tree-optimization/86066
15939 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
15940 for BIT_INSERT_EXPR stores.
15942 2018-06-06 Richard Biener <rguenther@suse.de>
15944 PR tree-optimization/86062
15945 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
15946 component refs ontop
15947 of to be offsetted base.
15949 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
15951 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
15952 to be static and remove check on interrupt attribute name.
15954 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
15956 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
15957 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
15959 2018-06-05 Steve Ellcey <sellcey@cavium.com>
15962 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
15964 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
15965 Remove second argument, change how error is called.
15966 (aarch64_layout_arg): Remove second argument from
15967 aarch64_err_no_fpadvsimd call.
15968 (aarch64_init_cumulative_args): Ditto.
15969 (aarch64_gimplify_va_arg_expr): Ditto.
15970 * config/aarch64/aarch64.md (mov<mode>): Ditto.
15972 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
15974 * config/i386/i386.md (simple_return_indirect_internal): New expander.
15975 (*simple_return_indirect_internal<mode>): Rename from
15976 simple_return_indirect_internal. Use W mode iterator.
15977 (rstorssp): New expander.
15978 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
15979 (clrssbsy): New expander.
15980 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
15982 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
15984 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
15986 (cmse_check_pointed_object): Likewise.
15988 2018-06-05 Martin Liska <mliska@suse.cz>
15990 PR gcov-profile/47618
15991 * doc/invoke.texi: Document how -fprofile-dir format
15994 2018-06-05 Richard Biener <rguenther@suse.de>
15996 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
15997 removal pretend DOM info isn't available so we do not update
15998 it and only remove edges, not dominated blocks. Actually free
15999 DOM info in case we removed something. Remove unreachable blocks.
16000 (mfb_keep_latches): Work with either DOM info or marked backedges.
16001 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
16002 first. Mark backedges if DOM info isn't available.
16003 (Re-)compute DOM info after cleanup_control_flow_pre.
16005 2018-06-05 Richard Biener <rguenther@suse.de>
16007 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
16008 (locus_discrim_hasher::hash): Adjust.
16009 (locus_discrim_hasher::equal): Likewise.
16010 (next_discriminator_for_locus): Work on line directly.
16011 (same_line_p): Pass in expanded locus1 as well.
16012 (assign_discriminators): Avoid redundant location expansions.
16014 2018-06-05 Richard Biener <rguenther@suse.de>
16016 PR tree-optimization/86046
16017 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
16018 if required after clearing TREE_ADDRESSABLE.
16020 2018-06-05 Richard Biener <rguenther@suse.de>
16022 PR tree-optimization/86047
16023 * tree-ssa-loop.c (for_each_index): Glob handling of all
16024 decls and constants and really handle all of them.
16026 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16029 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
16030 qualifier_void_pointer and qualifier_const_void_pointer.
16031 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
16032 (arm_init_builtins): Handle the above.
16033 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
16034 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
16035 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
16038 2018-06-05 Martin Liska <mliska@suse.cz>
16040 * auto-profile.c (read_autofdo_file): Do not use
16041 gcov_ctr_summary struct.
16042 (afdo_callsite_hot_enough_for_early_inline): Likewise.
16043 * coverage.c (struct counts_entry): Likewise.
16044 (read_counts_file): Read just single summary entry.
16045 (get_coverage_counts): Use gcov_summary struct.
16046 * coverage.h (get_coverage_counts): Likewise.
16047 * gcov-dump.c (dump_working_sets): Likewise.
16048 (tag_summary): Dump just single summary.
16049 * gcov-io.c (gcov_write_summary): Write just histogram
16051 (gcov_read_summary): Read just single summary.
16052 (compute_working_sets): Use gcov_summary struct.
16053 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
16054 of GCOV_COUNTERS_SUMMABLE.
16055 (GCOV_COUNTERS_SUMMABLE): Remove.
16056 (GCOV_FIRST_VALUE_COUNTER): Replace with
16057 GCOV_COUNTER_V_INTERVAL.
16058 (struct gcov_ctr_summary): Remove.
16059 (struct gcov_summary): Directly use fields of former
16061 (compute_working_sets): Use gcov_summary struct.
16062 * gcov.c (read_count_file): Do not use ctrs fields.
16063 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
16065 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
16067 * profile.c: Likewise.
16068 * profile.h: Likewise.
16070 2018-06-05 Martin Liska <mliska@suse.cz>
16072 PR gcov-profile/84846
16073 * gcov.c (output_lines): Print working directory only
16074 in intermediate format.
16076 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
16078 * config/s390/s390-builtin-types.def: Add void function type.
16079 * config/s390/s390-builtins.def: Use the function type for the
16082 2018-06-04 Jim Wilson <jimw@sifive.com>
16084 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
16086 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
16087 and maybe_eh_return. Change regno to unsigned int. Use new args to
16088 handle EH_RETURN_DATA_REGNO registers properly.
16089 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
16090 (riscv_expand_epilogue): Update comment. Change argument name and
16091 type. Update code to use new name and type. Pass new args to
16092 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
16094 * config/riscv/riscv.md (NORMAL_RETURN): New.
16095 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
16096 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
16097 (eh_return): Call gen_eh_return_internal and emit barrier.
16098 (eh_return_internal): Call riscv_expand_epilogue.
16100 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
16102 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
16103 bit_insertion field and declare can_be_merged_into method.
16104 (merged_store_group::can_be_merged_into): New method.
16105 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
16106 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
16107 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
16109 2018-06-04 Richard Biener <rguenther@suse.de>
16111 PR tree-optimization/85955
16112 * builtins.c (fold_builtin_sincos): Convert pointers to
16113 destination to appropriate type before dereferencing.
16115 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
16117 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
16119 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
16121 * expr.c (expand_expr_real_1): Force the operand into memory if
16122 its TYPE_MODE is BLKmode and if there is no integer mode for
16123 the number of bits being extracted.
16125 2018-06-04 Jakub Jelinek <jakub@redhat.com>
16129 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
16130 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
16132 2018-06-04 Richard Biener <rguenther@suse.de>
16134 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
16135 (cleanup_tree_cfg_noloop): ... single caller. Do
16136 start_recording_case_labels later.
16138 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
16140 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
16141 to _IMMINTRIN_H_INCLUDED.
16142 * config/i386/pconfigintrin.h: Ditto.
16143 * config/i386/waitpkgintrin.h: Ditto.
16144 * config/i386/immintrin.h: Add includes for sgxintrin.h,
16145 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
16146 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
16147 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
16148 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
16149 waitpkgintrin.h and cldemoteintrin.h.
16151 2018-06-04 Richard Biener <rguenther@suse.de>
16153 PR tree-optimization/86038
16154 * tracer.c (find_best_successor): Check probability for
16155 being initialized, bail out if not.
16157 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
16160 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
16161 of bits to ignore when comparing architectures.
16163 2018-06-04 Jakub Jelinek <jakub@redhat.com>
16165 PR tree-optimization/69615
16166 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
16167 maximum or minimum of the type, try to merge it also as if
16168 range1 is + [-, x - 1] or + [x + 1, -].
16171 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
16173 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
16175 PR tree-optimization/86034
16176 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
16177 the unsigned bitfield type in a bit insertion sequence if it does not
16178 have a larger precision than the bitfield size.
16179 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
16181 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
16183 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
16185 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
16187 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
16188 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
16189 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
16190 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
16192 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
16194 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
16195 Disable -fdelete-null-pointer-checks for ELF toolchain.
16197 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
16198 Kito Cheng <kito.cheng@gmail.com>
16200 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
16201 (nds32le-*-*, nds32be-*-*): Integrate checking process.
16202 (nds32*-*-*): Add glibc and uclibc conditions.
16203 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
16204 (TARGET_EXCEPT_UNWIND_INFO): Define.
16205 * config/nds32/elf.h: New file.
16206 * config/nds32/linux.h: New file.
16207 * config/nds32/nds32-elf.opt: New file.
16208 * config/nds32/nds32-linux.opt: New file.
16209 * config/nds32/nds32-fp-as-gp.c
16210 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
16211 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
16213 (nds32_asm_file_end): Ditto.
16214 (nds32_print_operand): Ditto.
16215 (nds32_insert_attributes): Ditto.
16216 (nds32_init_libfuncs): New function.
16217 (TARGET_HAVE_TLS): Define.
16218 (TARGET_INIT_LIBFUNCS): Define.
16219 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
16221 (TARGET_ELF): Apply different mcmodel setting.
16222 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
16223 been migrated into elf.h and linux.h files.
16224 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
16225 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
16226 (mcmodel): The content has been migrated into nds32-elf.opt and
16227 nds32-linux.opt files.
16228 * config/nds32/t-elf: New file.
16229 * config/nds32/t-linux: New file.
16231 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
16232 Shiva Chen <shiva0217@gmail.com>
16234 * config/nds32/constants.md (unspec_volatile_element): Add
16235 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
16236 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
16238 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
16239 (make_pass_nds32_fp_as_gp): Declare.
16240 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
16242 (nds32_asm_function_end_prologue): Remove unused asm output.
16243 (nds32_asm_function_begin_epilogue): Remove unused asm output.
16244 (nds32_asm_file_start): Output necessary fp_as_gp information.
16245 (nds32_option_override): Adjust register usage.
16246 (nds32_expand_prologue): Consider fp_as_gp situation.
16247 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
16248 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
16251 (simple_return): Ditto.
16252 (omit_fp_begin): Output special directive for fp_as_gp.
16253 (omit_fp_end): Output special directive for fp_as_gp.
16254 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
16255 mforbid-fp-as-gp): New options.
16257 2018-06-01 Mark Wielaard <mark@klomp.org>
16259 * dwarf2out.c (dwarf2out_finish): Remove generation of
16260 DW_AT_loclists_base.
16262 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
16264 * gimple-ssa-store-merging.c: Include gimple-fold.h.
16265 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
16266 (struct merged_store_group): Add bit_insertion field.
16267 (dump_char_array): Use standard hexadecimal format.
16268 (merged_store_group::merged_store_group): Set bit_insertion to false.
16269 (merged_store_group::apply_stores): Use optimal buffer size. Deal
16270 with BIT_INSERT_EXPR stores. Move up code updating the mask and
16271 also print the mask in the dump file.
16272 (pass_store_merging::gate): Minor tweak.
16273 (imm_store_chain_info::coalesce_immediate): Fix wrong association
16274 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
16275 stores with INTEGER_CST stores.
16276 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
16277 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
16278 and use it throughout. Generate bit insertion sequences if need be.
16279 (pass_store_merging::process_store): Remove redundant condition.
16280 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
16282 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
16284 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
16285 the 128-bit floating point types. Fix function comment.
16287 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16289 * config/aarch64/aarch64-simd.md
16290 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
16292 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
16295 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
16297 PR tree-optimization/85989
16298 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
16300 (backprop::intersect_uses): Check it when deciding whether this
16301 is a backedge reference.
16302 (backprop::process_block): Add each phi to m_visited_phis
16303 after visiting it, then clear it at the end.
16305 2018-06-01 Richard Biener <rguenther@suse.de>
16307 * tree-vectorizer.h (vect_dr_stmt): New function.
16308 (vect_get_load_cost): Adjust.
16309 (vect_get_store_cost): Likewise.
16310 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
16311 Use vect_dr_stmt instead of DR_SMTT.
16312 (vect_record_base_alignments): Likewise.
16313 (vect_calculate_target_alignment): Likewise.
16314 (vect_compute_data_ref_alignment): Likewise and make static.
16315 (vect_update_misalignment_for_peel): Likewise.
16316 (vect_verify_datarefs_alignment): Likewise.
16317 (vector_alignment_reachable_p): Likewise.
16318 (vect_get_data_access_cost): Likewise. Pass down
16319 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
16320 (vect_get_peeling_costs_all_drs): Likewise.
16321 (vect_peeling_hash_get_lowest_cost): Likewise.
16322 (vect_enhance_data_refs_alignment): Likewise.
16323 (vect_find_same_alignment_drs): Likewise.
16324 (vect_analyze_data_refs_alignment): Likewise.
16325 (vect_analyze_group_access_1): Likewise.
16326 (vect_analyze_group_access): Likewise.
16327 (vect_analyze_data_ref_access): Likewise.
16328 (vect_analyze_data_ref_accesses): Likewise.
16329 (vect_vfa_segment_size): Likewise.
16330 (vect_small_gap_p): Likewise.
16331 (vectorizable_with_step_bound_p): Likewise.
16332 (vect_prune_runtime_alias_test_list): Likewise.
16333 (vect_analyze_data_refs): Likewise.
16334 (vect_supportable_dr_alignment): Likewise.
16335 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
16336 (vect_gen_prolog_loop_niters): Likewise.
16337 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
16338 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
16340 (vect_recog_mask_conversion_pattern): Likewise.
16341 (vect_try_gather_scatter_pattern): Likewise.
16342 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
16343 to vect_get_store_cost.
16344 (vect_get_store_cost): Get stmt_info instead of DR.
16345 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
16346 (vect_get_load_cost): Get stmt_info instead of DR.
16348 2018-06-01 Richard Biener <rguenther@suse.de>
16350 PR middle-end/86017
16351 * gimple-fold.c (var_decl_component_p): Also allow offsetted
16352 vars wrapped in MEM_REFs.
16354 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
16356 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
16357 Fix subreg tests so that we only return a choice between
16358 GENERAL_REGS and FP_REGS if the original classes included both.
16360 2018-06-01 Richard Biener <rguenther@suse.de>
16363 * tree-ssa-structalias.c (get_function_part_constraint):
16364 Handle NULL fi->decl.
16365 (find_func_aliases_for_call): Properly handle indirect
16366 fi from direct call.
16367 (find_func_clobbers): Likewise.
16368 (ipa_pta_execute): Likewise.
16369 (create_variable_info_for): For functions that are ifunc_resolver
16370 resolve to a varinfo that contains the result of the resolver call.
16371 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
16374 2018-05-31 Michael Collison <michael.collison@arm.com>
16376 * config/aarch64/aarch64.md:
16377 (*fix_to_zero_extenddfdi2): New pattern.
16378 * gcc.target/aarch64/fix_extend1.c: New testcase.
16380 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
16382 PR middle-end/78809
16383 PR middle-end/83026
16384 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
16385 and BUILT_IN_STRNCMP_EQ.
16386 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
16387 BUILT_IN_STRNCMP_EQ.
16388 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
16389 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
16390 (gimple_fold_builtin): Likewise.
16391 * tree-ssa-strlen.c (compute_string_length): New function.
16392 (determine_min_obsize): New function.
16393 (handle_builtin_string_cmp): New function to handle calls to
16394 string compare functions.
16395 (strlen_optimize_stmt): Add handling to builtin string compare
16397 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
16398 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
16399 * tree.c (build_common_builtin_nodes): Add new defines of
16400 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
16402 2018-05-31 Jakub Jelinek <jakub@redhat.com>
16405 * bb-reorder.c (pass_partition_blocks::gate): Return false for
16406 functions with naked attribute.
16408 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
16410 * config/i386/sse.md (avx_vec_concat<mode>):
16411 Substitute concat_tg_mode mode attribute with xtg_mode.
16412 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
16413 (concat_tg_mode): Remove mode attribute.
16415 2018-05-31 Martin Sebor <msebor@redhat.com>
16418 * calls.c (alloc_max_size): Correct a logic error/typo.
16419 Treat excessive arguments as infinite. Warn for invalid arguments.
16420 * doc/invoke.texi (-Walloc-size-larger-than): Update.
16422 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
16425 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
16426 and movx for Haswell.
16428 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
16429 Cesar Philippidis <cesar@codesourcery.com>
16431 PR middle-end/85879
16432 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
16433 when emitting error on private/firstprivate reductions.
16434 * omp-low.c (lower_omp_target): Avoid reference-type processing
16435 on pointers for firstprivate clause.
16437 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
16439 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
16442 * config/aarch64/aarch64-simd.md
16443 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
16444 (aarch64_ld1_x3_<mode>): Likewise
16445 (aarch64_st1x2<VALLDIF:mode>): Likewise
16446 (aarch64_st1_x2_<mode>): Likewise
16447 (aarch64_st1x3<VALLDIF:mode>): Likewise
16448 (aarch64_st1_x3_<mode>): Likewise
16449 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
16450 (vld1_s8_x3): Likewise.
16451 (vld1_u16_x3): Likewise.
16452 (vld1_s16_x3): Likewise.
16453 (vld1_u32_x3): Likewise.
16454 (vld1_s32_x3): Likewise.
16455 (vld1_u64_x3): Likewise.
16456 (vld1_s64_x3): Likewise.
16457 (vld1_f16_x3): Likewise.
16458 (vld1_f32_x3): Likewise.
16459 (vld1_f64_x3): Likewise.
16460 (vld1_p8_x3): Likewise.
16461 (vld1_p16_x3): Likewise.
16462 (vld1_p64_x3): Likewise.
16463 (vld1q_u8_x3): Likewise.
16464 (vld1q_s8_x3): Likewise.
16465 (vld1q_u16_x3): Likewise.
16466 (vld1q_s16_x3): Likewise.
16467 (vld1q_u32_x3): Likewise.
16468 (vld1q_s32_x3): Likewise.
16469 (vld1q_u64_x3): Likewise.
16470 (vld1q_s64_x3): Likewise.
16471 (vld1q_f16_x3): Likewise.
16472 (vld1q_f32_x3): Likewise.
16473 (vld1q_f64_x3): Likewise.
16474 (vld1q_p8_x3): Likewise.
16475 (vld1q_p16_x3): Likewise.
16476 (vld1q_p64_x3): Likewise.
16477 (vst1_s64_x2): Likewise.
16478 (vst1_u64_x2): Likewise.
16479 (vst1_f64_x2): Likewise.
16480 (vst1_s8_x2): Likewise.
16481 (vst1_p8_x2): Likewise.
16482 (vst1_s16_x2): Likewise.
16483 (vst1_p16_x2): Likewise.
16484 (vst1_s32_x2): Likewise.
16485 (vst1_u8_x2): Likewise.
16486 (vst1_u16_x2): Likewise.
16487 (vst1_u32_x2): Likewise.
16488 (vst1_f16_x2): Likewise.
16489 (vst1_f32_x2): Likewise.
16490 (vst1_p64_x2): Likewise.
16491 (vst1q_s8_x2): Likewise.
16492 (vst1q_p8_x2): Likewise.
16493 (vst1q_s16_x2): Likewise.
16494 (vst1q_p16_x2): Likewise.
16495 (vst1q_s32_x2): Likewise.
16496 (vst1q_s64_x2): Likewise.
16497 (vst1q_u8_x2): Likewise.
16498 (vst1q_u16_x2): Likewise.
16499 (vst1q_u32_x2): Likewise.
16500 (vst1q_u64_x2): Likewise.
16501 (vst1q_f16_x2): Likewise.
16502 (vst1q_f32_x2): Likewise.
16503 (vst1q_f64_x2): Likewise.
16504 (vst1q_p64_x2): Likewise.
16505 (vst1_s64_x3): Likewise.
16506 (vst1_u64_x3): Likewise.
16507 (vst1_f64_x3): Likewise.
16508 (vst1_s8_x3): Likewise.
16509 (vst1_p8_x3): Likewise.
16510 (vst1_s16_x3): Likewise.
16511 (vst1_p16_x3): Likewise.
16512 (vst1_s32_x3): Likewise.
16513 (vst1_u8_x3): Likewise.
16514 (vst1_u16_x3): Likewise.
16515 (vst1_u32_x3): Likewise.
16516 (vst1_f16_x3): Likewise.
16517 (vst1_f32_x3): Likewise.
16518 (vst1_p64_x3): Likewise.
16519 (vst1q_s8_x3): Likewise.
16520 (vst1q_p8_x3): Likewise.
16521 (vst1q_s16_x3): Likewise.
16522 (vst1q_p16_x3): Likewise.
16523 (vst1q_s32_x3): Likewise.
16524 (vst1q_s64_x3): Likewise.
16525 (vst1q_u8_x3): Likewise.
16526 (vst1q_u16_x3): Likewise.
16527 (vst1q_u32_x3): Likewise.
16528 (vst1q_u64_x3): Likewise.
16529 (vst1q_f16_x3): Likewise.
16530 (vst1q_f32_x3): Likewise.
16531 (vst1q_f64_x3): Likewise.
16532 (vst1q_p64_x3): Likewise.
16534 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
16536 * config/msp430/msp430.c (msp430_output_labelref): Prepend
16537 user_label_prefix to name.
16539 * tree-core.h: Update comment about the format of NAME string
16540 passed to handler in attribute_spec.
16542 * config/msp430/msp430.md: Remove erroneous subreg expression from
16543 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
16544 zero_extend{q,h}isi2.
16546 2018-05-30 Borislav Petkov <bp@suse.de>
16548 * doc/extend.texi: Document some architecture specific
16549 constraints and sort entries.
16551 2018-05-30 Martin Sebor <msebor@redhat.com>
16553 PR middle-end/85369
16554 * builtins.c (expand_builtin_stpcpy_1): New function.
16555 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
16556 only if the former succeeds.
16558 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
16560 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
16563 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
16565 * doc/invoke.texi (-flinker-output): Document
16567 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
16569 * passes.c (ipa_write_summaries): Only modify statements if body
16571 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
16572 incrementally linking.
16573 (ipa_passes): Likewise.
16574 * lto-cgraph.c (lto_output_node): When incrementally linking do not
16575 pass down resolution info.
16576 * common.opt (flag_incremental_link): Update info.
16577 * gcc.c (plugin specs): Turn flinker-output=* to
16578 -plugin-opt=-linker-output-known
16579 * toplev.c (compile_file): Also cut compilation when doing incremental
16581 * flag-types. (enum lto_partition_model): Add
16582 LTO_LINKER_OUTPUT_NOLTOREL.
16583 (invoke.texi): Add -flinker-output docs.
16584 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
16585 link same way as WPA; do not stream in dead initializers.
16587 * dwarf2out.c (dwarf2out_die_ref_for_decl,
16588 darf2out_register_external_decl): Support incremental link.
16590 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
16592 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
16594 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
16596 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
16597 it down to simple_object_copy_lto_debug_sections.
16598 (run_gcc): Determine incremental LTO link time and configure
16599 lto1 into non-wpa mode, disable renaming of debug sections.
16601 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
16603 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
16604 descriptions of various incorrectly documented functions.
16606 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
16609 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
16610 address check not strict.
16612 2018-05-30 Richard Biener <rguenther@suse.de>
16614 PR tree-optimization/85964
16615 * tracer.c (better_p): Drop initialized count check, we only
16616 call the function with initialized counts now.
16617 (find_best_successor): Do find a best edge if one
16618 has uninitialized count.
16619 (find_best_predecessor): Likewise. Do BB frequency check only
16620 if count is initialized.
16622 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
16624 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
16625 (aarch64_ldrstr_offset_compare): New.
16626 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
16627 load/store orderings.
16628 (aarch64_gen_adjusted_ldpstp): Likewise.
16630 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
16632 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
16633 Check for subset of GENERAL_REGS and FP_REGS.
16634 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
16637 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
16639 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
16640 and wi::to_poly_offset. Add the current offset and then check
16641 whether the sum fits, rather than using an unchecked addition of
16642 a checked term. Check for a shwi rather than a uhwi.
16643 * expr.c (get_bit_range): Use tree_to_poly_uint64.
16644 (store_constructor): Use poly_int_tree_p.
16645 (expand_expr_real_1): Likewise.
16646 * function.c (assign_temp): Likewise.
16647 * fold-const.c (const_binop): Use poly_int_tree_p and
16648 wi::to_poly_offset.
16649 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
16651 * ipa-icf-gimple.c (func_checker::compare_operand): Use
16652 to_poly_offset for MEM offsets.
16653 * ipa-icf.c (sem_variable::equals): Likewise.
16654 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
16655 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
16656 wi::to_poly_offset for BIT_FIELD_REF offsets.
16657 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
16658 wi::to_poly_offset.
16659 * var-tracking.c (emit_note_insn_var_location): Use
16660 tree_to_poly_uint64.
16662 2018-05-29 Jim Wilson <jimw@sifive.com>
16664 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
16666 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
16669 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
16670 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
16672 (sse4_1_round<mode>2): Use nonimmediate_operand
16673 for operand 1 predicate.
16675 2018-05-29 Martin Sebor <msebor@redhat.com>
16676 Richard Biener <rguenther@suse.de>
16679 * calls.c (get_size_range): Call determine_value_range instead
16680 of get_value_range..
16681 * tree-vrp.h (determine_value_range): Declared new function.
16682 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
16684 2018-05-29 Richard Biener <rguenther@suse.de>
16686 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
16687 sure to use non-pattern stmts for get_earlier_stmt arguments.
16688 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
16689 called on pattern stmts.
16690 (get_later_stmt): Likewise.
16692 2018-05-29 Martin Liska <mliska@suse.cz>
16694 PR gcov-profile/85759
16695 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
16698 2018-05-29 Jakub Jelinek <jakub@redhat.com>
16700 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
16702 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
16706 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
16707 VEC_PACK_FLOAT_EXPR): New tree codes.
16708 * tree-pretty-print.c (op_code_prio): Handle
16709 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
16710 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
16711 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
16712 * tree-inline.c (estimate_operator_cost): Likewise.
16713 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
16714 * fold-const.c (const_binop): Likewise.
16715 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
16716 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
16717 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
16718 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
16719 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
16720 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
16721 * expr.c (expand_expr_real_2): Likewise.
16722 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
16723 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
16724 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
16726 * optabs.c (expand_widen_pattern_expr): For
16727 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
16728 sign from result type rather than operand's type.
16729 (expand_binop_directly): For vec_packu_float_optab and
16730 vec_packs_float_optab allow result type to be different from operand's
16732 * optabs-tree.c (optab_for_tree_code): Handle
16733 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
16734 VEC_PACK_FLOAT_EXPR. Formatting fixes.
16735 * tree-vect-generic.c (expand_vector_operations_1): Handle
16736 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
16737 VEC_PACK_FLOAT_EXPR.
16738 * tree-vect-stmts.c (supportable_widening_operation): Handle
16740 (supportable_narrowing_operation): Handle FLOAT_EXPR.
16741 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
16742 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
16743 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
16744 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
16746 (vec_pack<floatprefix>_float_<mode>): New expander.
16747 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
16749 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
16750 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
16751 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
16752 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
16753 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
16755 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
16756 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
16757 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
16758 VEC_PACK_FLOAT_EXPR): Document.
16760 2018-05-29 Richard Biener <rguenther@suse.de>
16762 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
16764 (stmt_vec_info_vec): Make pointer.
16765 (init_stmt_vec_info_vec): Remove.
16766 (free_stmt_vec_info_vec): Likewise.
16767 (set_stmt_vec_info_vec): New function.
16768 (free_stmt_vec_infos): Likewise.
16769 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
16770 (set_vinfo_for_stmt): Likewise.
16771 (get_earlier_stmt): Likewise.
16772 (get_later_stmt): Likewise.
16773 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
16774 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
16775 (vec_info::~vec_info): Free stmt_vec_infos.
16776 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
16777 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
16778 (pass_slp_vectorize::execute): Likewise.
16779 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
16780 (free_stmt_vec_info_vec): Likewise.
16781 (set_stmt_vec_info_vec): New function.
16782 (free_stmt_vec_infos): Likewise.
16783 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
16784 the global stmt_vec_info_vec.
16785 * tree-parloops.c (gather_scalar_reductions): Use
16786 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
16789 2018-05-29 Richard Biener <rguenther@suse.de>
16791 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
16793 2018-05-29 Martin Liska <mliska@suse.cz>
16794 David Malcolm <dmalcolm@redhat.com>
16796 * vec.c (test_reverse): New.
16797 (vec_c_tests): Add new test.
16798 * vec.h (vl_ptr>::reverse): New function.
16800 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
16802 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
16804 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
16807 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
16809 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
16811 2018-05-28 Richard Biener <rguenther@suse.de>
16813 PR tree-optimization/85933
16814 * tree-vect-data-refs.c (vect_record_base_alignments): Only
16815 look at stmts marked as vectorizable.
16817 2018-05-28 Richard Biener <rguenther@suse.de>
16819 PR tree-optimization/85934
16820 * tree-vect-generic.c (expand_vector_operations_1): Hoist
16821 vector boolean check before scalar optimization.
16823 2018-05-28 Jakub Jelinek <jakub@redhat.com>
16825 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
16828 2018-05-28 Mark Wielaard <mark@klomp.org>
16830 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
16831 if it is an expression containing a minus sign.
16833 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
16835 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
16837 2018-05-27 Paul Koning <ni1d@arrl.net>
16839 * config/pdp11/pdp11.md (truncsihi2): Remove.
16841 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
16842 Chung-Ju Wu <jasonwucj@gmail.com>
16844 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
16846 (unaligned_store_dw): Ditto.
16847 * config/nds32/nds32-memory-manipulation.c
16848 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
16849 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
16850 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
16851 (emit_setmem_word_loop): Rename to ...
16852 (emit_setmem_doubleword_loop): ... this.
16853 (nds32_gen_dup_4_byte_to_word_value): New function.
16854 (nds32_gen_dup_8_byte_to_double_word_value): New function.
16855 (nds32_expand_setmem_loop): Refine implementation.
16856 (nds32_expand_setmem_loop_v3m): Ditto.
16857 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
16860 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
16862 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
16864 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
16866 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
16867 (nds32_init_machine_status): Initialize machine->attr_naked_p and
16868 machine->attr_no_prologue_p.
16869 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
16870 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
16871 (nds32_expand_epilogue): Consider attr_naked_p.
16872 (nds32_expand_epilogue_v3pop): Likewise.
16873 (nds32_can_use_return_insn): Likewise.
16874 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
16875 attr_no_prologue_p fields.
16876 * config/nds32/nds32.opt (mret-in-naked-func): New option.
16878 2018-05-27 Jakub Jelinek <jakub@redhat.com>
16881 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
16883 * config/i386/sse.md
16884 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
16886 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
16888 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
16890 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
16892 (*<floatsuffix>floatv2div2sf2): Rename to ...
16893 (*float<floatunssuffix>v2div2sf2): ... this.
16894 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
16895 (float<floatunssuffix>v2div2sf2_mask): ... this.
16896 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
16897 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
16898 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
16900 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
16902 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
16904 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
16906 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
16908 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
16910 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
16911 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
16912 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
16913 gen_ufix_truncv8dfv8si2.
16914 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
16915 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
16916 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
16917 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
16918 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
16919 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
16920 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
16921 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
16923 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
16927 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
16929 2018-05-25 Jim Wilson <jimw@sifive.com>
16931 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
16932 * config/riscv/riscv.c (struct machine_function): Add
16933 interrupt_handler_p and attribute_checked_p fields.
16934 (riscv_attribute_table): Add interrupt.
16935 (riscv_interrupt_type_p): New.
16936 (riscv_save_reg_p): Save extra regs for interrupt handler.
16937 (riscv_use_save_libcall): Return false for interrupt handler.
16938 (riscv_first_stack_step): Add forward declaration.
16939 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
16940 for interrupt handler with large frame. Use it for saved reg list.
16941 (riscv_expand_prologue): Move flag_stack_usage_info support to
16942 eliminate duplication.
16943 (riscv_expand_epilogue): Generate mret for interrupt handler.
16944 (riscv_epilogue_uses): New.
16945 (riscv_can_use_return_insn): Return false for interrupt handler.
16946 (riscv_function_ok_for_sibcall): Likewise.
16947 (riscv_set_current_function): Add interrupt handler support.
16948 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
16949 * config/riscv/riscv.md (UNSPECV_MRET): New.
16951 (riscv_frflags, riscv_fsflags): Use tab after opcode.
16953 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
16955 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
16957 PR tree-optimization/85712
16958 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
16959 this candidate has already been replaced in-situ by a copy.
16961 2018-05-25 Jason Merrill <jason@redhat.com>
16963 PR c++/80485 - inline function non-zero address.
16964 * symtab.c (nonzero_address): Check DECL_COMDAT.
16966 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
16969 * config/alpha/alpha.md (ashlsi3): New insn pattern.
16970 (*ashlsi_se): Rename from *ashldi_se. Define as sign
16971 extension of SImode operation. Use const123_operand predicate.
16972 (*saddsi_1): Remove.
16973 (*saddl_se_1): Ditto.
16974 (*ssubsi_1): Ditto.
16975 (*ssubl_se_1): Ditto.
16976 * config/alpha/predicates.md (const123_operand): New predicate.
16977 * config/alpha/constraints.md (P): Use IN_RANGE.
16979 2018-05-25 Richard Biener <rguenther@suse.de>
16981 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
16983 (ref_maybe_used_by_stmt_p): Likewise.
16984 (stmt_may_clobber_ref_p): Likewise.
16985 (stmt_may_clobber_ref_p_1): Likewise.
16986 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
16988 (ref_maybe_used_by_stmt_p): Likewise.
16989 (stmt_may_clobber_ref_p): Likewise.
16990 (stmt_may_clobber_ref_p_1): Likewise.
16991 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
16992 the alias oracle to disambiguate DRs with stmts DR analysis
16994 (vect_analyze_data_refs): Do not give up on not analyzable
16995 DRs for BB vectorization. Remove code truncating the dataref
16998 2018-05-25 Jakub Jelinek <jakub@redhat.com>
17001 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17002 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
17003 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
17005 2018-05-25 Richard Biener <rguenther@suse.de>
17007 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
17008 function, combining stmt data ref gathering and fatal analysis
17010 (vect_analyze_data_refs): Remove now redudnant code and simplify.
17011 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
17012 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
17013 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
17014 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
17016 2018-05-25 Bin Cheng <bin.cheng@arm.com>
17018 PR tree-optimization/85720
17019 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
17020 SCC if all partitions are builtins.
17021 (version_loop_by_alias_check): New parameter. Generate cancelable
17022 runtime alias check if all partitions are builtins.
17023 (distribute_loop): Update call to above function.
17025 2018-05-25 Bin Cheng <bin.cheng@arm.com>
17027 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
17028 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
17029 (parm_default_def_partition_arg): Ditto.
17030 (set_parm_default_def_partition): Ditto.
17031 (get_parm_default_def_partitions): Ditto and make it static.
17032 (get_undefined_value_partitions): Ditto and make it static.
17033 (remove_ssa_form): Refactor call to init_var_map here.
17034 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
17035 computation for loop region.
17036 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
17037 (register_default_def): Delete.
17038 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
17039 (parm_default_def_partition_arg): Ditto.
17040 (set_parm_default_def_partition): Ditto.
17041 (get_parm_default_def_partitions): Ditto and make it static.
17042 (get_undefined_value_partitions): Ditto and make it static.
17043 (coalesce_with_default, coalesce_with_default): Update comment.
17044 (create_coalesce_list_for_region): New func factored out from
17045 create_outofssa_var_map.
17046 (populate_coalesce_list_for_outofssa): New func factored out from
17047 create_outofssa_var_map and coalesce_ssa_name.
17048 (create_outofssa_var_map): Delete.
17049 (coalesce_ssa_name): Refactor to support live range computation.
17050 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
17051 (get_parm_default_def_partitions): Delete.
17052 (get_undefined_value_partitions): Ditto.
17053 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
17054 computation for loop region.
17055 (new_tree_live_info, loe_visit_block): Ditto.
17056 (live_worklist, set_var_live_on_entry): Ditto.
17057 (calculate_live_on_exit, verify_live_on_entry): Ditto.
17058 * tree-ssa-live.h (struct _var_map): New fields.
17059 (init_var_map): Change decl.
17060 (region_contains_p): New.
17062 2018-05-25 Bin Cheng <bin.cheng@arm.com>
17064 * tree-ssa-live.h (live_merge_and_clear): Delete.
17066 2018-05-25 Richard Biener <rguenther@suse.de>
17069 * tree-dump.c (dequeue_and_dump): Remove access to removed
17070 operand 2 of a SWITCH_EXPR.
17072 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
17074 * doc/sourcebuild.texi (vect_double_cond_arith): Include
17075 multiplication and division.
17076 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
17077 (cond_udiv@var{m}, cond_umod@var{m}): Document.
17078 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
17079 (cond_udiv_optab, cond_umod_optab): New optabs.
17080 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
17081 (IFN_COND_RDIV): New internal functions.
17082 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
17083 TRUNC_MOD_EXPR and RDIV_EXPR.
17084 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
17085 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
17087 (SVE_INT_BINARY): Include mult.
17088 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
17089 (optab, sve_int_op): Handle mult.
17090 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
17092 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
17093 for SVE_INT_BINARY_SD.
17095 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
17097 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
17098 (optab, sve_int_op): Handle div and udiv.
17099 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
17100 for SVE_INT_BINARY_SD.
17101 (*<optab><mode>3): New insn for the same.
17103 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
17105 * tree-vect-patterns.c: Include predict.h.
17106 (vect_recog_divmod_pattern): Restrict check for division support
17107 to when optimizing for size.
17109 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
17111 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
17112 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
17113 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
17114 (gimple_match_op::set_op): Likewise.
17115 (gimple_resimplify4): Declare.
17116 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
17117 (expr::gen_transform): Likewise.
17118 (decision_tree::gen): Generate a simplification routine for 4 operands.
17119 * gimple-match-head.c (gimple_simplify): Add an overload for
17120 4 operands. In the top-level function, handle up to 4 call
17121 arguments and call gimple_resimplify4.
17122 (gimple_resimplify4): New function.
17123 (build_call_internal): Pass a fourth operand.
17124 (maybe_push_to_seq): Likewise.
17125 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
17126 Fold VEC_COND_EXPRs of an operation and a default value into
17127 an IFN_COND_* function if possible.
17128 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
17130 (SVE_COND_FP_BINARY): Include them.
17131 (optab, sve_fp_op): Handle them.
17132 (SVE_INT_BINARY_REV): New code iterator.
17133 (SVE_COND_FP_BINARY_REV): New int iterator.
17134 (commutative): New int attribute.
17135 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
17137 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
17139 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
17140 (*cond_<optab><mode>): New patterns for reversed operands.
17142 2018-05-25 Richard Biener <rguenther@suse.de>
17144 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
17145 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
17146 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
17147 (STMT_VINFO_GROUPED_ACCESS): Adjust.
17148 * tree-vect-data-refs.c (everywhere): Adjust users.
17149 * tree-vect-loop.c (everywhere): Likewise.
17150 * tree-vect-slp.c (everywhere): Likewise.
17151 * tree-vect-stmts.c (everywhere): Likewise.
17152 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
17154 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17156 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
17158 (gcc_cv_as_section_exclude): ... this.
17159 Try Solaris as #exclude syntax.
17160 * configure: Regenerate.
17161 * config.in: Regenerate.
17162 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
17164 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
17165 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
17167 * varasm.c (default_elf_asm_named_section): Don't check if
17168 HAVE_GAS_SECTION_EXCLUDE is defined.
17170 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
17172 * doc/md.texi: Update the documentation of the cond_* optabs
17173 to mention the new final operand. Fix GET_MODE_NUNITS call.
17174 Describe the scalar case too.
17175 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
17176 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
17178 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
17179 (get_conditional_internal_fn): Update comment.
17180 * tree-vect-loop.c (vectorizable_reduction): Pass the original
17181 accumulator value as a final argument to conditional functions.
17182 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
17183 a define_expand and add an "else" operand. Assert for now that
17184 the else operand is equal to operand 2. Use SVE_INT_BINARY and
17185 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
17186 (*cond_<optab><mode>): New patterns.
17187 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
17188 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
17189 (UNSPEC_COND_EOR): Delete.
17190 (optab): Remove associated mappings.
17191 (SVE_INT_BINARY): New code iterator.
17192 (sve_int_op): Remove int attribute and add "minus" to the code
17194 (SVE_COND_INT_OP): Delete.
17195 (SVE_COND_FP_OP): Rename to...
17196 (SVE_COND_FP_BINARY): ...this.
17198 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
17200 * optabs.c (can_reuse_operands_p): New function.
17201 (maybe_legitimize_operands): Try to reuse the results for
17204 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
17206 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
17207 Add {q} suffix to insn mnemonic.
17209 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
17211 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
17212 (msp430_warn_func_return): New.
17214 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
17216 * fold-const.c (tree_nonzero_bits): New function.
17217 * fold-const.h (tree_nonzero_bits): Likewise.
17218 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
17219 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
17221 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
17225 * varasm.c (assemble_alias): Check ifunc_resolver only on
17228 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
17231 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
17232 when memory input operand is handled.
17234 2018-05-24 Luis Machado <luis.machado@linaro.org>
17236 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
17238 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
17240 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
17242 * match.pd: Delay FMA folds until after vectorization.
17244 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
17247 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
17248 address check not strict.
17250 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
17252 * gimple-match.h (gimple_match_op): New class.
17253 (mprts_hook): Replace parameters with a gimple_match_op *.
17254 (maybe_build_generic_op): Likewise.
17255 (gimple_simplified_result_is_gimple_val): Replace parameters with
17256 a const gimple_match_op *.
17257 (gimple_simplify): Replace code_helper * and tree * parameters with
17258 a gimple_match_op * parameter.
17259 (gimple_resimplify1): Replace code_helper *, tree and tree *
17260 parameters with a gimple_match_op * parameter.
17261 (gimple_resimplify2): Likewise.
17262 (gimple_resimplify3): Likewise.
17263 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
17264 parameters with a gimple_match_op * parameter.
17265 * gimple-match-head.c (gimple_simplify): Change prototypes of
17266 auto-generated functions to take a gimple_match_op * instead of
17267 separate code_helper * and tree * parameters. Make the same
17268 change in the top-level overload and update calls to the
17269 gimple_resimplify routines. Update calls to the auto-generated
17270 functions and to maybe_push_res_to_seq in the publicly-facing
17271 operation-specific gimple_simplify overloads.
17272 (gimple_match_op::MAX_NUM_OPS): Define.
17273 (gimple_resimplify1): Replace rcode and ops with a single res_op
17274 parameter. Update call to gimple_simplify.
17275 (gimple_resimplify2): Likewise.
17276 (gimple_resimplify3): Likewise.
17277 (mprts_hook): Replace parameters with a gimple_match_op *.
17278 (maybe_build_generic_op): Likewise.
17279 (build_call_internal): Replace type, nargs and ops with
17280 a gimple_match_op *.
17281 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
17282 with a single gimple_match_op *. Update calls to mprts_hook,
17283 build_call_internal and gimple_simplified_result_is_gimple_val.
17284 Factor out code that is common to the tree_code and combined_fn cases.
17285 * genmatch.c (expr::gen_transform): Replace tem_code and
17286 tem_ops with a gimple_match_op called tem_op. Update calls
17287 to the gimple_resimplify functions and maybe_push_res_to_seq.
17288 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
17289 res_ops. Update call to the gimple_resimplify functions.
17290 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
17291 (decision_tree::gen): Make the functions take a gimple_match_op *
17292 called res_op instead of separate res_code and res_ops parameters.
17293 Update call accordingly.
17294 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
17295 and ops with a single res_op parameter. Update calls to
17296 maybe_build_generic_op and maybe_push_res_to_seq.
17297 (fold_stmt_1): Update calls to gimple_simplify and
17298 replace_stmt_with_simplification.
17299 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
17300 and gimple_simplified_result_is_gimple_val.
17301 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
17303 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
17304 with a gimple_match_op *.
17305 (vn_nary_build_or_lookup): Likewise. Update call to
17306 vn_nary_build_or_lookup_1.
17307 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
17308 gimple_match_op *. Update calls to the gimple_resimplify routines
17309 and to gimple_simplified_result_is_gimple_val.
17310 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
17311 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
17312 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
17313 (visit_nary_op): Likewise.
17314 (visit_reference_op_load): Likewise.
17316 2018-05-23 Luis Machado <luis.machado@linaro.org>
17318 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
17319 modifier for printing the step amount.
17321 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
17324 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
17327 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
17329 * doc/sourcebuild.texi (Endianness): New subsubsection.
17331 2018-05-23 Luis Machado <luis.machado@linaro.org>
17333 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17334 <prefetch_dynamic_strides>: New const bool field.
17335 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17336 prefetch_dynamic_strides.
17337 (exynosm1_prefetch_tune): Likewise.
17338 (thunderxt88_prefetch_tune): Likewise.
17339 (thunderx_prefetch_tune): Likewise.
17340 (thunderx2t99_prefetch_tune): Likewise.
17341 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
17343 (aarch64_override_options_internal): Update to set
17344 PARAM_PREFETCH_DYNAMIC_STRIDES.
17345 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
17346 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
17347 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
17348 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
17349 prefetch-dynamic-strides setting.
17351 2018-05-23 Luis Machado <luis.machado@linaro.org>
17353 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17354 <minimum_stride>: New const int field.
17355 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17356 minimum_stride field defaulting to -1.
17357 (exynosm1_prefetch_tune): Likewise.
17358 (thunderxt88_prefetch_tune): Likewise.
17359 (thunderx_prefetch_tune): Likewise.
17360 (thunderx2t99_prefetch_tune): Likewise.
17361 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
17362 <default_opt_level>: Set to 3.
17363 (aarch64_override_options_internal): Update to set
17364 PARAM_PREFETCH_MINIMUM_STRIDE.
17365 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
17366 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
17367 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
17368 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
17369 stride is constant and is below the minimum stride threshold.
17371 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17373 * config/arm/arm-cpus.in (mode26): Delete.
17374 (armv4): Delete mode26 reference.
17375 * config/arm/arm.c (arm_configure_build_target): Delete use of
17378 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
17380 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
17382 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
17383 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
17385 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
17386 (floatunsdidf2): Ditto.
17388 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
17390 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
17391 (fixuns_trunc<mode>si2_avx512f): Ditto.
17392 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
17393 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
17394 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
17396 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
17398 PR rtl-optimization/79985
17399 * df-scan.c (df_insn_refs_collect): Remove special case for
17400 global registers and asm statements.
17402 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
17404 * extend.texi (Global Register Variables): Rewrite the bullet list.
17405 Note that the register is available for allocation. Note that access
17406 via inline asm must use constraints. Add note about async-signal
17407 handlers. Remove paragraph about automagic register selection.
17409 2018-05-23 Richard Biener <rguenther@suse.de>
17411 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
17412 of fixed offset from memset VN.
17414 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
17416 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
17417 first_interp field.
17418 (alloc_cand_and_find_basis): Initialize first_interp field.
17419 (slsr_process_mul): Modify first_interp field.
17420 (slsr_process_add): Likewise.
17421 (slsr_process_cast): Modify first_interp field for each new
17423 (slsr_process_copy): Likewise.
17424 (dump_candidate): Dump first_interp field.
17425 (replace_mult_candidate): Process all interpretations, not just
17427 (replace_rhs_if_not_dup): Likewise.
17428 (replace_one_candidate): Likewise.
17430 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
17432 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
17434 (aarch64_needs_frame_chain): New function.
17435 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
17437 2018-05-23 Sudakshina Das <sudi.das@arm.com>
17440 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
17441 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
17442 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
17443 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
17444 as true for strict-align.
17445 (aarch64_can_inline_p): Perform checks even when callee has no
17446 attributes to check for strict alignment.
17447 * doc/extend.texi (AArch64 Function Attributes): Document
17449 * doc/invoke.texi: (AArch64 Options): Likewise.
17451 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
17453 PR tree-optimization/85853
17454 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
17455 the handling of the root of the node to...
17456 (vect_slp_analyze_node_operations_1): ...this new function,
17457 and run the whole thing with the child nodes' def types
17458 set according to their SLP node's def type.
17460 2018-05-23 Richard Biener <rguenther@suse.de>
17462 PR middle-end/85874
17463 * tree-data-ref.c (create_runtime_alias_checks): Defer
17464 and ignore overflow warnings.
17466 2018-05-23 Yury Gribov <tetra2005@gmail.com>
17468 PR tree-optimization/85822
17469 * tree-vrp.c (is_masked_range_test): Fix handling of negative
17472 2018-05-23 Richard Biener <rguenther@suse.de>
17474 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
17475 memset constants via native_interpret_expr.
17477 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
17480 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
17482 (cgraph_node::create_alias): Likewise.
17483 (cgraph_node::get_availability): Check ifunc_resolver instead
17484 of looking up ifunc attribute.
17485 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
17486 * varasm.c (do_assemble_alias): Likewise.
17487 (assemble_alias): Likewise.
17488 (default_binds_local_p_3): Likewise.
17489 * cgraph.h (cgraph_node): Add ifunc_resolver.
17490 (cgraph_node::only_called_directly_or_aliased_p): Return false
17491 for IFUNC resolver.
17492 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
17494 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
17495 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
17496 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
17497 instead of looking up ifunc attribute.
17499 2018-05-22 Luis Machado <luis.machado@linaro.org>
17501 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
17503 2018-05-22 Martin Sebor <msebor@redhat.com>
17505 PR middle-end/85359
17506 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
17507 only when expasion succeeds.
17508 (expand_builtin_strcmp): Same.
17509 (expand_builtin_strncmp): Same.
17511 2018-05-22 Martin Sebor <msebor@redhat.com>
17513 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
17515 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
17516 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17518 * config/aarch64/aarch64-ldpstp.md: Replace uses of
17519 aarch64_mem_pair_operand with memory_operand and delete operand swapping
17521 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
17522 Add check for legitimate_address.
17523 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
17524 (aarch64_swap_ldrstr_operands): New.
17525 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
17528 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
17529 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17531 * config/aarch64/aarch64.md: New patterns to generate stp
17533 (store_pair_sw, store_pair_dw): New patterns to generate stp for
17534 single words and double words.
17535 (load_pair_sw, load_pair_dw): Likewise.
17536 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
17538 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
17540 * config/aarch64/aarch64-ldpstp.md: Modify peephole
17541 for different mode ldpstp and add peephole for merged zero stores.
17542 Likewise for loads.
17543 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
17545 (aarch64_gen_store_pair): Rename calls to match new patterns.
17546 (aarch64_gen_load_pair): Rename calls to match new patterns.
17547 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
17548 (load_pair<DREG:mode><DREG2:mode>): ... This.
17549 (store_pair<mode>): Rename to...
17550 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
17551 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
17552 New mode iterators.
17553 (V_INT_EQUIV): Handle SImode.
17554 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
17557 2018-05-22 Martin Sebor <msebor@redhat.com>
17560 * calls.c (maybe_warn_nonstring_arg): Use string length to set
17561 or ajust the presumed bound on an operation to avoid unnecessary
17564 2018-05-22 Martin Sebor <msebor@redhat.com>
17566 PR tree-optimization/85826
17567 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
17568 assuming that a DECL necesarily has a constant size.
17570 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
17572 PR middle-end/85862
17573 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
17575 2018-05-22 Richard Biener <rguenther@suse.de>
17577 PR tree-optimization/85834
17578 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
17579 non-constant and non-zero memset arguments.
17581 2018-05-22 Martin Liska <mliska@suse.cz>
17584 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
17586 2018-05-22 Richard Biener <rguenther@suse.de>
17588 PR tree-optimization/85863
17589 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
17590 comparisons when vectype is specified.
17591 (vectorizable_condition): Do not specify vectype for
17592 vect_is_simple_cond when SLP vectorizing.
17594 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
17597 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
17598 define __ibm128 as long double.
17599 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
17600 as a distinct type when IEEE 128-bit support is enabled.
17601 (init_float128_ieee): Fix up conversions between IFmode and IEEE
17602 128-bit types to use the correct functions.
17603 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
17604 convert between 128-bit floating point types that have different
17605 modes but the same representation, instead of using gen_lowpart to
17607 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
17609 (IFKF_reg): New attributes to give the register constraints for
17611 (extend<mode>tf2_internal): New insns to mark an explicit
17612 conversion between 128-bit floating point types that have a
17613 different mode but share the same representation.
17615 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
17617 PR tree-optimization/85814
17618 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
17619 a null return from get_strinfo when unsharing the next
17620 strinfo in the chain.
17622 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
17625 * varasm.c (weak_finish): Clean up weak_decls.
17627 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17629 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
17630 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
17631 UNSPEC_UADALP values.
17632 * config/aarch64/iterators.md (ABAL): New int iterator.
17635 (sur): Add mappings for the above.
17636 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
17638 (aarch64_<sur>abal<mode>_4): Likewise.
17639 (aarch64_<sur>adalp<mode>_3): Likewise.
17640 (<sur>sadv16qi): New define_expand.
17642 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
17644 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
17645 (*movdf_internal): Ditto.
17646 (*rcpsf2_sse): Ditto.
17647 (*rsqrtsf2_sse): Ditto.
17648 (*sqrt<mode>2_sse): Ditto.
17650 2018-05-21 Tamar Christina <tamar.christina@arm.com>
17652 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
17654 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
17655 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
17656 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
17657 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
17659 * config/aarch64/arm_neon.h: Likewise.
17660 * config/aarch64/iterators.md (VQ_I): New.
17662 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
17664 * config.gcc: Add arc/t-multilib-linux to tmake_file for
17666 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
17669 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
17671 * config/nds32/constraints.md (S): New constraint.
17672 * config/nds32/nds32.md (call_internal): Use constraint S.
17673 (call_value_internal): Likewise.
17674 (sibcall_internal): Likewise.
17675 (sibcall_value_internal): Likewise.
17677 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
17678 Chung-Ju Wu <jasonwucj@gmail.com>
17680 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
17681 into consideration.
17683 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
17684 Chung-Ju Wu <jasonwucj@gmail.com>
17686 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
17687 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
17688 (nds32_rtx_costs_impl): Simplify.
17689 (nds32_address_cost_impl): Simplify.
17690 (nds32_init_rtx_costs): New function.
17691 (nds32_rtx_costs_speed_prefer): Likewise.
17692 (nds32_rtx_costs_size_prefer): Likewise.
17693 (nds32_address_cost_speed_prefer): Likewise.
17694 (nds32_address_cost_speed_fwprop): Likewise.
17695 (nds32_address_cost_size_prefer): Likewise.
17696 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
17697 * config/nds32/nds32.c (nds32_option_override): Use
17698 nds32_init_rtx_costs function.
17700 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
17702 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
17703 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
17704 (TARGET_PIPELINE_N8): Likewise.
17705 (TARGET_PIPELINE_N10): Likewise.
17706 (TARGET_PIPELINE_N13): Likewise.
17707 (TARGET_PIPELINE_GRAYWOLF): Likewise.
17709 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
17711 * config/nds32/nds32-fpu.md: Update copyright year.
17713 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
17715 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
17717 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
17719 * config/nds32/nds32.c
17720 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
17721 * config/nds32/nds32.opt (minline-asm-r15): New option.
17723 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
17725 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
17727 * config/nds32/nds32.md (abssi2): New pattern.
17729 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
17731 * config/i386/i386.md (rex64namesuffix): New mode attribute.
17732 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
17733 Merge insn pattern from sse_cvtsi2ss<round_name> and
17734 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
17735 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
17736 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
17737 using SWI48 mode iterator.
17738 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
17739 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
17740 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
17741 pattern from sse_cvttss2si<round_saeonly_name>
17742 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
17743 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
17744 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
17745 using SWI48 mode iterator.
17746 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
17747 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
17748 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
17749 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
17750 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
17751 using SWI48 mode iterator.
17752 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
17753 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
17754 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
17755 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
17756 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
17757 SWI48 mode iterator.
17758 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
17759 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
17760 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
17761 pattern from sse_cvttsd2si<round_saeonly_name>
17762 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
17764 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
17766 * config/nds32/nds32-md-auxiliary.c
17767 (nds32_valid_smw_lwm_base_p): Refine.
17768 (nds32_output_smw_single_word): Refine.
17769 (nds32_output_smw_double_word): New.
17770 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
17772 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
17774 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
17775 (nds32_output_stack_pop): Refine.
17776 (nds32_expand_unaligned_load): Refine.
17777 (nds32_expand_unaligned_store): Refine.
17779 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
17780 Chung-Ju Wu <jasonwucj@gmail.com>
17782 * config/nds32/constants.md: Add TP_REGNUM constant.
17783 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
17784 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
17786 * config/nds32/nds32-doubleword.md: Consider flag_pic.
17787 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
17788 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
17789 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
17790 and PIC code generation.
17791 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
17793 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
17795 * config/nds32/nds32.md: Support TLS and PIC.
17796 * config/nds32/nds32.c: Support TLS and PIC.
17797 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
17798 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
17801 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
17803 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
17804 mode with E_ prefix.
17806 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
17807 Chung-Ju Wu <jasonwucj@gmail.com>
17809 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
17810 * config/nds32/nds32-md-auxiliary.c
17811 (symbolic_reference_mentioned_p): New.
17812 (nds32_legitimize_ict_address): New.
17813 (nds32_expand_ict_move): New.
17814 (nds32_indirect_call_referenced_p): New.
17815 (nds32_symbol_binds_local_p): Delete.
17816 (nds32_long_call_p): Modify.
17817 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
17818 * config/nds32/nds32-protos.h
17819 (symbolic_reference_mentioned_p): Declare.
17820 (nds32_legitimize_ict_address): Declare.
17821 (nds32_expand_ict_move): Declare.
17822 (nds32_indirect_call_referenced_p): Declare.
17823 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
17824 (nds32_relax_group): Use nds32_ict_const_p as condition.
17825 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
17826 (nds32_asm_file_start): Output ict_model directive in asm code.
17827 (nds32_legitimate_address_p): Consider indirect call.
17828 (nds32_print_operand): Consider indirect call.
17829 (nds32_print_operand_address): Consider indirect call.
17830 (nds32_insert_attributes): Handle "indirect_call" attribute.
17831 (TARGET_LEGITIMATE_ADDRESS_P): Define.
17832 (TARGET_LEGITIMATE_CONSTANT_P): Define.
17833 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
17834 (TARGET_DELEGITIMIZE_ADDRESS): Define.
17835 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
17836 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
17837 (TARGET_ICT_MODEL_SMALL): Define.
17838 (TARGET_ICT_MODEL_LARGE): Define.
17839 * config/nds32/nds32.md (movsi): Consider ict model.
17840 (call, call_value): Consider ict model.
17841 (sibcall, sibcall_value): Consider ict model.
17842 * config/nds32/nds32.opt (mict-model): New option.
17843 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
17846 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
17847 Monk Chiang <sh.chiang04@gmail.com>
17848 Jim Wilson <jimw@sifive.com>
17850 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
17851 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
17852 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
17853 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
17854 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
17855 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
17856 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
17857 compute save_libcall_adjustment properly.
17858 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
17859 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
17860 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
17861 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
17862 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
17863 (ABI_SPEC): Handle mabi=ilp32e.
17864 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
17865 (RVE): Add RVE mask.
17866 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
17867 <-march>: Add rv32e as an example.
17869 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
17872 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
17873 (intra_create_variable_infos): Handle C++ constructors.
17875 2018-05-18 Martin Liska <mliska@suse.cz>
17877 * passes.def: Remove a redundant pass.
17879 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
17882 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
17884 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17886 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
17888 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
17890 (armv2, armv2a, armv3, armv3m): Delete architectures.
17891 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
17892 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
17893 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
17895 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
17896 (*mulsidi3adddi): Likewise.
17897 (mulsidi3): Likewise.
17898 (*mulsidi3_nov6): Likewise.
17899 (umulsidi3): Likewise.
17900 (umulsidi3_nov6): Likewise.
17901 (umaddsidi4): Likewise.
17902 (*umulsidi3adddi): Likewise.
17903 (smulsi3_highpart): Likewise.
17904 (*smulsi3_highpart_nov6): Likewise.
17905 (umulsi3_highpart): Likewise.
17906 (*umulsi3_highpart_nov6): Likewise.
17907 * config/arm/arm.h (arm_arch3m): Delete.
17908 * config/arm/arm.c (arm_arch3m): Delete.
17909 (arm_option_override_internal): Update armv3-related comment.
17910 (arm_configure_build_target): Delete use of isa_bit_mode32.
17911 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
17912 (arm_rtx_costs_internal): Delete check of arm_arch3m.
17913 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
17914 (mulsa3): Likewise.
17915 (mulusa3): Likewise.
17916 * config/arm/arm-protos.h (arm_arch3m): Delete.
17917 * config/arm/arm-tables.opt: Regenerate.
17918 * config/arm/arm-tune.md: Likewise.
17919 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
17920 deleted architectures.
17922 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17924 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
17925 (armv5t, armv5te): New features.
17926 (ARMv5, ARMv5e): Delete fgroups.
17927 (ARMv5t, ARMv5te): Adjust for above changes.
17928 (ARMv6m): Likewise.
17929 (armv5, armv5e): Delete arches.
17930 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
17932 (*call_reg_arm): Likewise.
17933 (*call_value_reg_armv5): Likewise.
17934 (*call_value_reg_arm): Likewise.
17935 (*call_symbol): Likewise.
17936 (*call_value_symbol): Likewise.
17937 (*sibcall_insn): Likewise.
17938 (*sibcall_value_insn): Likewise.
17939 (clzsi2): Likewise.
17940 (prefetch): Likewise.
17941 (define_split and define_peephole2 dependent on arm_arch5):
17943 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
17945 (TARGET_ARM_QBIT): Likewise.
17946 (TARGET_DSP_MULTIPLY): Likewise.
17947 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
17948 (arm_arch5, arm_arch5e): Delete.
17949 (arm_arch5t, arm_arch5te): Declare.
17950 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
17951 (arm_arch5t): Declare.
17952 (arm_option_reconfigure_globals): Update for the above.
17953 (arm_options_perform_arch_sanity_checks): Update comment, replace
17954 use of arm_arch5 with arm_arch5t.
17955 (use_return_insn): Likewise.
17956 (arm_emit_call_insn): Likewise.
17957 (output_return_instruction): Likewise.
17958 (arm_final_prescan_insn): Likewise.
17959 (arm_coproc_builtin_available): Likewise.
17960 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
17961 arm_arch5e with arm_arch5t and arm_arch5te.
17962 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
17963 (arm_arch5t, arm_arch5te): Declare.
17964 * config/arm/arm-tables.opt: Regenerate.
17965 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
17966 * config/arm/t-multilib: Likewise.
17967 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
17968 instead of arm_arch5.
17969 (*call_reg_thumb1): Likewise.
17970 (*call_value_reg_thumb1_v5): Likewise.
17971 (*call_value_reg_thumb1): Likewise.
17972 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
17974 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
17976 2018-05-18 Martin Liska <mliska@suse.cz>
17978 PR gcov-profile/84846
17979 * doc/gcov.texi: Document -t option of gcov tool.
17981 2018-05-18 Martin Liska <mliska@suse.cz>
17983 PR gcov-profile/84846
17984 * gcov.c (print_usage): Add new -t option.
17985 (process_args): Handle the option.
17986 (generate_results): Use stdout as output when requested by
17989 2018-05-18 Martin Liska <mliska@suse.cz>
17991 PR gcov-profile/84846
17992 * coverage.c (coverage_init): Write PWD to .gcno file.
17993 * doc/gcov.texi: Document how working directory is printed.
17994 * gcov-dump.c (dump_gcov_file): Print PWD.
17995 * gcov.c (output_intermediate_file): Likewise.
17996 (read_graph_file): Read PWD string.
17997 (output_lines): Print PWD.
17999 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18001 PR middle-end/85817
18002 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
18003 for retval and return false if all args to phi are zero.
18005 2018-05-18 Richard Biener <rguenther@suse.de>
18007 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
18009 (evrp_dom_walker::before_dom_children): Call it.
18011 2018-05-18 Richard Biener <rguenther@suse.de>
18013 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
18014 results when processing array refs with variable index.
18016 2018-05-18 Toon Moene <toon@moene.org>
18018 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
18019 directly after that of -floop-interchange. Indicate that both
18020 options are enabled by default when specifying -O3.
18022 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18024 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
18025 iterator. Delete separate integer-mode vec_set<mode> expander.
18026 (aarch64_simd_vec_setv2di): Delete.
18027 (vec_setv2di): Delete.
18028 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
18029 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
18030 the "w, r" alternative.
18032 2018-05-18 Martin Liska <mliska@suse.cz>
18034 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
18035 * tree-pass.h (make_pass_lower_switch_O0): New function.
18036 * tree-switch-conversion.c (node_has_low_bound): Remove.
18037 (node_has_high_bound): Likewise.
18038 (node_is_bounded): Likewise.
18039 (class pass_lower_switch): Make it a template type and create
18041 (pass_lower_switch::execute): Add template argument.
18042 (make_pass_lower_switch): New function.
18043 (make_pass_lower_switch_O0): New function.
18044 (do_jump_if_equal): Remove.
18045 (emit_case_nodes): Simplify to just handle all 3 cases and leave
18046 all the hard work to tree optimization passes.
18048 2018-05-18 Martin Liska <mliska@suse.cz>
18050 * dbgcnt.c (limit_low): Renamed from limit.
18051 (limit_high): New variable.
18052 (dbg_cnt_is_enabled): Check for upper limit.
18053 (dbg_cnt): Adjust dumping.
18054 (dbg_cnt_set_limit_by_index): Add new argument for high
18056 (dbg_cnt_set_limit_by_name): Likewise.
18057 (dbg_cnt_process_single_pair): Parse new format.
18058 (dbg_cnt_process_opt): Use strtok.
18059 (dbg_cnt_list_all_counters): Remove 'value' and add
18061 * doc/invoke.texi: Document changes.
18063 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
18065 * doc/sourcebuild.texi (scalar_all_fma): Document.
18066 * tree.def (FMA_EXPR): Delete.
18067 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
18068 * internal-fn.c (ternary_direct): New macro.
18069 (expand_ternary_optab_fn): Likewise.
18070 (direct_ternary_optab_supported_p): Likewise.
18071 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
18072 * builtins.c (fold_builtin_fma): Delete.
18073 (fold_builtin_3): Don't call it.
18074 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
18075 * expr.c (expand_expr_real_2): Likewise.
18076 * fold-const.c (operand_equal_p): Likewise.
18077 (fold_ternary_loc): Likewise.
18078 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
18079 * gimple.c (DEFTREECODE): Likewise.
18080 * gimplify.c (gimplify_expr): Likewise.
18081 * optabs-tree.c (optab_for_tree_code): Likewise.
18082 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18083 * tree-eh.c (operation_could_trap_p): Likewise.
18084 (stmt_could_throw_1_p): Likewise.
18085 * tree-inline.c (estimate_operator_cost): Likewise.
18086 * tree-pretty-print.c (dump_generic_node): Likewise.
18087 (op_code_prio): Likewise.
18088 * tree-ssa-loop-im.c (stmt_cost): Likewise.
18089 * tree-ssa-operands.c (get_expr_operands): Likewise.
18090 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
18091 * fold-const-call.h (fold_fma): Delete.
18092 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
18093 CFN_FNMA and CFN_FNMS.
18094 (fold_fma): Delete.
18095 * genmatch.c (combined_fn): New enum.
18096 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
18097 (commutative_op): New function.
18098 (commutate): Use it. Handle more than 2 operands.
18099 (dt_operand::gen_gimple_expr): Use commutative_op.
18100 (parser::parse_expr): Allow :c to be used with non-binary
18101 operators if the commutative operand is known.
18102 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
18103 CFN_FMS, CFN_FNMA and CFN_FNMS.
18104 (backprop::process_assign_use): Remove FMA_EXPR handling.
18105 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
18106 (gen_hsa_fma): New function.
18107 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
18108 IFN_FNMA and IFN_FNMS.
18109 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
18110 * gimple-fold.h (follow_all_ssa_edges): Declare.
18111 * gimple-fold.c (follow_all_ssa_edges): New function.
18112 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
18113 gimple_build interface and use follow_all_ssa_edges to fold the result.
18114 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
18115 instead of checking for optabs directly.
18116 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
18117 rather than FMA_EXPRs.
18118 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
18119 call to IFN_FMA instead of an FMA_EXPR.
18121 2018-05-17 Jim Wilson <jimw@sifive.com>
18123 * expr.c (do_tablejump): When converting index to Pmode, if we have a
18124 sign extended promoted subreg, and the range does not have the sign bit
18125 set, then do a sign extend.
18127 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
18128 test, check for sign extended subreg and/or constant operands, and
18129 do a sign extend in that case.
18131 2018-05-17 Steve Ellcey <sellcey@cavium.com>
18133 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
18134 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
18136 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
18137 Change logics_shift_reg to logics_shift_imm.
18138 (thunderx2t99_fp_loadpair_basic): Delete.
18139 (thunderx2t99_fp_storepair_basic): Delete.
18140 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
18141 (thunderx2t99_asimd_polynomial): Delete.
18142 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
18143 and neon_fp_mul_d_scalar_q.
18144 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
18145 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
18146 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
18147 (thunderx2t99_asimd_lut): Add missing tbl types.
18148 (thunderx2t99_asimd_ext): Delete.
18149 (thunderx2t99_asimd_load1_1_mult): Delete.
18150 (thunderx2t99_asimd_load1_2_mult): Delete.
18151 (thunderx2t99_asimd_load1_ldp): New.
18152 (thunderx2t99_asimd_load1): New.
18153 (thunderx2t99_asimd_load2): Add missing *load2* types.
18154 (thunderx2t99_asimd_load3): New.
18155 (thunderx2t99_asimd_load4): New.
18156 (thunderx2t99_asimd_store1_1_mult): Delete.
18157 (thunderx2t99_asimd_store1_2_mult): Delete.
18158 (thunderx2t99_asimd_store2_mult): Delete.
18159 (thunderx2t99_asimd_store2_onelane): Delete.
18160 (thunderx2t99_asimd_store_stp): New.
18161 (thunderx2t99_asimd_store1): New.
18162 (thunderx2t99_asimd_store2): New.
18163 (thunderx2t99_asimd_store3): New.
18164 (thunderx2t99_asimd_store4): New.
18166 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
18168 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
18169 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
18171 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
18172 Segher Boessenkool <segher@kernel.crashing.org>
18175 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
18178 2018-05-17 Richard Biener <rguenther@suse.de>
18180 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
18181 for pruning loop and prune defs feeding only already visited PHIs.
18183 2018-05-17 Richard Biener <rguenther@suse.de>
18185 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
18187 2018-05-17 Bin Cheng <bin.cheng@arm.com>
18188 Richard Biener <rguenther@suse.de>
18190 PR tree-optimization/85793
18191 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
18192 for VMAT_ELEMENTWISE.
18194 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
18196 * internal-fn.h (lookup_internal_fn): Declare
18197 * internal-fn.c (lookup_internal_fn): New function.
18198 * gimple.c (gimple_build_call_from_tree): Handle calls to
18199 internal functions.
18200 * gimple-pretty-print.c (dump_gimple_call): Print "." before
18201 internal function names.
18202 * tree-pretty-print.c (dump_generic_node): Likewise.
18203 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
18205 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
18207 * gimple-fold.h (gimple_build): Make the function forms take
18208 combined_fn rather than built_in_function.
18209 (gimple_simplify): Likewise.
18210 * gimple-match-head.c (gimple_simplify): Likewise.
18211 * gimple-fold.c (gimple_build): Likewise.
18212 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
18213 rather than gimple_build_call_internal.
18214 (get_initial_defs_for_reduction): Likewise.
18215 (vect_create_epilog_for_reduction): Likewise.
18216 (vectorizable_live_operation): Likewise.
18218 2018-05-17 Martin Liska <mliska@suse.cz>
18220 * gimple-ssa-sprintf.c (format_directive): Do not use
18221 space in between 'G_' and '('.
18223 2018-05-17 Jakub Jelinek <jakub@redhat.com>
18226 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
18227 even if the mask is not all ones.
18230 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
18232 (ix86_gimple_fold_builtin): Likewise.
18235 * config/i386/i386.c: Include tree-vector-builder.h.
18236 (ix86_vector_shift_count): New function.
18237 (ix86_fold_builtin): Fold shift builtins by scalar count.
18238 (ix86_gimple_fold_builtin): Likewise.
18240 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
18241 _mm512_setzero): New intrinsics.
18243 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
18244 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18246 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
18247 code generation for cases where splatting a value is not useful.
18248 * simplify-rtx.c (simplify_ternary_operation): Simplify
18249 vec_merge across a vec_duplicate and a paradoxical subreg forming
18250 a vector mode to a vec_concat.
18252 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
18254 * config.gcc: Support "goldmont-plus".
18255 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
18257 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18258 PROCESSOR_GOLDMONT_PLUS.
18259 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
18260 (processor_target_table): Add "goldmont-plus".
18261 (PTA_GOLDMONT_PLUS): Define.
18262 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
18263 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
18264 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
18265 (fold_builtin_cpu): Add "goldmont-plus".
18266 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
18267 (ix86_option_override_internal): Add "goldmont-plus".
18268 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
18269 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
18270 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
18271 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
18273 2018-05-17 Richard Biener <rguenther@suse.de>
18275 PR tree-optimization/85757
18276 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
18277 remove defs that only feed that PHI from further processing.
18279 2018-05-16 Jim Wilson <jimw@sifive.com>
18281 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
18283 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
18285 2018-05-16 Mark Wielaard <mark@klomp.org>
18287 * dwarf2out.c (count_index_strings): New function.
18288 (output_indirect_strings): Call count_index_strings and generate
18289 header for dwarf_version >= 5.
18291 2018-05-16 Mark Wielaard <mark@klomp.org>
18293 * dwarf2out.c (dwarf_FORM): New function.
18294 (set_indirect_string): Use dwarf_FORM.
18295 (reset_indirect_string): Likewise.
18296 (size_of_die): Likewise.
18297 (value_format): Likewise.
18298 (output_die): Likewise.
18299 (add_skeleton_AT_string): Likewise.
18300 (output_macinfo_op): Likewise.
18301 (index_string): Likewise.
18302 (output_index_string_offset): Likewise.
18303 (output_index_string): Likewise.
18304 (count_index_strings): Likewise.
18306 2018-05-16 Carl Love <cel@us.ibm.com>
18308 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
18309 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
18311 2018-05-16 Martin Jambor <mjambor@suse.cz>
18313 * ipa-prop.c (ipa_free_all_edge_args): Remove.
18314 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
18316 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
18318 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
18319 (fnma<mode>4): Likewise.
18320 (fms<mode>4): Likewise.
18321 (fnms<mode>4): Likewise.
18322 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
18323 (aarch64_fnma<mode>4): Likewise.
18324 (aarch64_fms<mode>4): Likewise.
18325 (aarch64_fnms<mode>4): Likewise.
18326 (aarch64_fnmadd<mode>4): Likewise.
18328 2018-05-16 Jason Merrill <jason@redhat.com>
18330 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
18332 2018-05-16 Richard Biener <rguenther@suse.de>
18334 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
18335 (dump_stmt_cost): Declare.
18336 (add_stmt_cost): Dump cost we add.
18337 (add_stmt_costs): New function.
18338 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
18339 No longer exported.
18340 (vect_analyze_stmt): Adjust prototype.
18341 (vectorizable_condition): Likewise.
18342 (vectorizable_live_operation): Likewise.
18343 (vectorizable_reduction): Likewise.
18344 (vectorizable_induction): Likewise.
18345 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
18346 cost vector to pass to vectorizable_ and record afterwards.
18347 (vect_model_reduction_cost): Take cost vector argument and adjust.
18348 (vect_model_induction_cost): Likewise.
18349 (vectorizable_reduction): Likewise.
18350 (vectorizable_induction): Likewise.
18351 (vectorizable_live_operation): Likewise.
18352 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
18353 SLP_TREE_NUMBER_OF_VEC_STMTS.
18354 (vect_analyze_slp_cost_1): Remove.
18355 (vect_analyze_slp_cost): Likewise.
18356 (vect_slp_analyze_node_operations): Take visited args and
18357 a target cost vector. Avoid processing already visited stmt sets.
18358 (vect_slp_analyze_operations): Use a local cost vector to gather
18359 costs and register those of non-discarded instances.
18360 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
18361 (vect_schedule_slp_instance): Remove copying of
18362 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
18364 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
18365 adding cost. Record cost entry location.
18366 (vect_prologue_cost_for_slp_op): Function to compute cost of
18367 a constant or invariant generated for SLP vect in the prologue,
18368 split out from vect_analyze_slp_cost_1.
18369 (vect_model_simple_cost): Make static. Adjust for SLP costing.
18370 (vect_model_promotion_demotion_cost): Likewise.
18371 (vect_model_store_cost): Likewise, make static.
18372 (vect_model_load_cost): Likewise.
18373 (vectorizable_bswap): Add cost vector arg and adjust.
18374 (vectorizable_call): Likewise.
18375 (vectorizable_simd_clone_call): Likewise.
18376 (vectorizable_conversion): Likewise.
18377 (vectorizable_assignment): Likewise.
18378 (vectorizable_shift): Likewise.
18379 (vectorizable_operation): Likewise.
18380 (vectorizable_store): Likewise.
18381 (vectorizable_load): Likewise.
18382 (vectorizable_condition): Likewise.
18383 (vectorizable_comparison): Likewise.
18384 (can_vectorize_live_stmts): Likewise.
18385 (vect_analyze_stmt): Likewise.
18386 (vect_transform_stmt): Adjust calls to vectorizable_*.
18387 * tree-vectorizer.c: Include gimple-pretty-print.h.
18388 (dump_stmt_cost): New function.
18390 2018-05-16 Richard Biener <rguenther@suse.de>
18392 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
18393 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
18394 * tree-ssa-dse.c: Include tree-ssa-loop.h.
18395 (check_name): New callback.
18396 (dse_classify_store): Track cycles via a visited bitmap of PHI
18397 defs and simplify handling of in-loop and across loop dead stores
18398 and properly fail for loop-variant refs. Handle byte-tracking with
18399 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
18402 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
18404 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
18405 (vect_get_mask_type_for_stmt): Likewise.
18406 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
18408 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
18409 to determine the statement's vector type and the vector type that
18410 should be used for calculating nunits. Deal with cases in which
18411 the type has to be deferred.
18412 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
18413 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
18414 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
18415 (vect_determine_vf_for_stmt): New functions, split out from...
18416 (vect_determine_vectorization_factor): ...here.
18417 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
18418 (vect_get_mask_type_for_stmt): New functions, split out from
18419 vect_determine_vectorization_factor.
18421 2018-05-16 Richard Biener <rguenther@suse.de>
18423 * tree-cfg.c (verify_gimple_assign_ternary): Properly
18424 verify the [VEC_]COND_EXPR embedded comparison.
18426 2018-05-15 Martin Sebor <msebor@redhat.com>
18428 PR tree-optimization/85753
18429 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
18430 RECORD_TYPE in addition to ARRAY_TYPE.
18432 2018-05-15 Martin Sebor <msebor@redhat.com>
18434 PR middle-end/85643
18435 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
18437 2018-05-15 Richard Biener <rguenther@suse.de>
18439 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
18440 add by_clobber_p one. Change algorithm to collect all defs
18441 representing uses we need to walk and try reducing them to
18442 a single one before failing.
18443 (dse_dom_walker::dse_optimize_stmt): Adjust.
18445 2018-05-13 Mark Wielaard <mark@klomp.org>
18447 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
18448 (size_of_loc_descr): Likewise.
18449 (output_loc_operands): Likewise.
18450 (output_loc_operands_raw): Likewise.
18451 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
18452 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
18453 (hash_loc_operands): Likewise.
18454 (compare_loc_operands): Likewise.
18456 2018-05-14 Mark Wielaard <mark@klomp.org>
18458 * dwarf2out.c (count_index_addrs): New function.
18459 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
18461 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18463 PR tree-optimization/83648
18464 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
18465 return value as malloc candidate.
18467 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18470 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
18471 param as true in call to suggest_attribute.
18473 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
18475 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
18476 -mreadonly-in-sdata.
18478 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18480 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
18482 (aarch64_crypto_aesd_fused): Likewise.
18484 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
18486 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
18487 (movsi_aarch64): Likewise.
18488 (load_pairsi): Likewise.
18489 (load_pairdi): Likewise.
18490 (store_pairsi): Likewise.
18491 (store_pairdi): Likewise.
18492 (load_pairsf): Likewise.
18493 (load_pairdf): Likewise.
18494 (store_pairsf): Likewise.
18495 (store_pairdf): Likewise.
18496 (zero_extend): Likewise.
18497 (trunc): Swap alternatives.
18498 (fcvt_target): Add '?' to prefer w over r.
18500 2018-05-14 Jakub Jelinek <jakub@redhat.com>
18503 * config/i386/i386.md: Disallow non-commutative arithmetics in
18504 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
18505 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
18506 in the peephole2 before it.
18508 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
18510 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
18511 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
18512 (ix86_handle_option): Handle -mcldemote.
18513 * config.gcc: New header.
18514 * config/i386/cldemoteintrin.h: New file.
18515 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
18516 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
18518 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18519 OPTION_MASK_ISA_CLDEMOTE.
18520 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
18521 (ix86_valid_target_attribute_inner_p): Ditto.
18522 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
18523 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
18524 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
18525 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
18526 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
18528 * config/i386/i386.opt: Add -mcldemote.
18529 * config/i386/x86intrin.h: New header.
18530 * doc/invoke.texi: Add -mcldemote.
18532 2018-05-14 Richard Biener <rguenther@suse.de>
18534 * doc/match-and-simplify.texi: Adjust :s documentation.
18536 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
18538 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
18539 intended memcpy size.
18540 (REORDER_45): Likewise.
18542 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
18544 * sort.cc: New file.
18545 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
18546 * vec.c (qsort_chk): Use gcc_qsort.
18547 * Makefile.in (OBJS-libcommon): Add sort.o.
18548 (build/sort.o): New target. Use it...
18549 (BUILD_RTL): ... here, and...
18550 (build/gencfn-macros): ... here, and...
18551 (build/genmatch): ... here.
18553 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
18554 Chung-Ju Wu <jasonwucj@gmail.com>
18556 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
18557 * config/nds32/nds32-graywolf.md: New file.
18558 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
18559 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
18561 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
18562 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
18563 * config/nds32/nds32.md (pipeline_model): Add graywolf.
18564 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
18565 * config/nds32/pipelines.md: Include n15 settings.
18567 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
18568 Chung-Ju Wu <jasonwucj@gmail.com>
18570 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
18571 * config/nds32/nds32-n13.md: New file.
18572 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
18573 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
18575 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
18576 * config/nds32/nds32.md (pipeline_model): Add n13.
18577 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
18578 * config/nds32/pipelines.md: Include n13 settings.
18580 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
18581 Chung-Ju Wu <jasonwucj@gmail.com>
18583 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
18584 * config/nds32/nds32-n10.md: New file.
18585 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
18586 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
18588 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
18589 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
18590 * config/nds32/nds32.md (pipeline_model): Add n10.
18591 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
18592 * config/nds32/pipelines.md: Include n10 settings.
18594 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
18595 Kito Cheng <kito.cheng@gmail.com>
18596 Chung-Ju Wu <jasonwucj@gmail.com>
18598 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
18599 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
18600 Add enum values for DSP extension instructions.
18601 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
18603 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
18604 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
18605 New code iterators.
18606 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
18607 * config/nds32/nds32-dspext.md: New file for DSP implementation.
18608 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
18609 * config/nds32/nds32-intrinsic.md: Likewise.
18610 * config/nds32/nds32_intrinsic.h: Likewise.
18611 * config/nds32/nds32-md-auxiliary.c: Likewise.
18612 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
18613 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
18614 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
18615 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
18616 * config/nds32/nds32-protos.h: New declarations for DSP extension.
18617 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
18618 TYPE_DMAC in switch statement.
18619 * config/nds32/nds32.c: New checking and implementation for DSP
18620 extension instructions.
18621 * config/nds32/nds32.h: Likewise.
18622 * config/nds32/nds32.md: Likewise.
18623 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
18624 * config/nds32/predicates.md: Implement new predicates for DSP
18627 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
18629 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
18630 Reformat alternatives and attributes so it is easier to identify
18631 which constraints/attributes go with which instruction.
18632 (mov<mode>_hardfloat32, FMOVE64): Likewise.
18633 (mov<mode>_softfloat32, FMOVE64): Likewise.
18634 (mov<mode>_hardfloat64, FMOVE64): Likewise.
18635 (mov<mode>_softfloat64, FMOVE64): Likewise.
18637 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
18639 * doc/extend.texi (PowerPC Built-in Functions): Rename this
18641 (Basic PowerPC Built-in Functions): The new name of the
18642 subsection previously known as "PowerPC Built-in Functions".
18643 (Basic PowerPC Built-in Functions Available on all Configurations):
18645 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
18646 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
18647 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
18648 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
18650 2018-05-11 Martin Jambor <mjambor@suse.cz>
18653 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
18656 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
18659 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
18661 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
18663 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
18664 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
18665 (ix86_handle_option): Handle -mwaitpkg.
18666 * config.gcc: New header.
18667 * config/i386/cpuid.h (bit_WAITPKG): New bit.
18668 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
18669 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
18671 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18672 OPTION_MASK_ISA_WAITPKG.
18673 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
18674 (ix86_option_override_internal): Add PTA_WAITPKG.
18675 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
18676 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
18677 IX86_BUILTIN_TPAUSE.
18678 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
18679 __builtin_ia32_umwait and __builtin_ia32_tpause.
18680 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
18681 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
18682 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
18683 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
18684 UNSPECV_TPAUSE): New.
18685 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
18686 * config/i386/i386.opt: Add -mwaitpkg.
18687 * config/i386/waitpkgintrin.h: New file.
18688 * config/i386/x86intrin.h: New header.
18689 * doc/invoke.texi: Add -mwaitpkg.
18691 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
18694 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
18696 (cortex-m0): Use armv6s-m isa.
18697 (cortex-m0plus): Likewise.
18698 (cortex-m1): Likewise.
18699 (cortex-m0.small-multiply): Likewise.
18700 (cortex-m0plus.small-multiply): Likewise.
18701 (cortex-m1.small-multiply): Likewise.
18703 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
18704 Jakub Jelinek <jakub@redhat.com>
18706 PR tree-optimization/85692
18707 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
18708 source permute as well.
18710 2018-05-11 Martin Liska <mliska@suse.cz>
18713 * doc/extend.texi: Document LLVM style format for no_sanitize
18716 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
18718 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
18719 mode_supports_vsx_dform_quad to mode_supports_dq_form.
18720 (mode_supports_vsx_dform_quad): Likewise.
18721 (mode_supports_vmx_dform): Move these functions to be next to the
18722 other mode_supports functions.
18723 (mode_supports_dq_form): Likewise.
18724 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
18725 mode_supports_dq_form.
18726 (reg_offset_addressing_ok_p): Likewise.
18727 (offsettable_ok_by_alignment): Likewise.
18728 (rs6000_legitimate_offset_address_p): Likewise.
18729 (legitimate_lo_sum_address_p): Likewise.
18730 (rs6000_legitimize_address): Likewise.
18731 (rs6000_legitimize_reload_address): Likewise.
18732 (rs6000_secondary_reload_inner): Likewise.
18733 (rs6000_preferred_reload_class): Likewise.
18734 (rs6000_output_move_128bit): Likewise.
18736 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
18738 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
18739 Generate SImode target register for null target.
18740 <case IX86_BUILTIN_XGETBV>: Ditto.
18741 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
18742 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
18744 2018-05-10 Carl Love <cel@us.ibm.com>
18746 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
18747 dcbtt and dcbtstt if operands[2] is 0.
18749 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
18752 * config/i386/sse.md (usadv64qi): New expander.
18754 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
18756 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
18757 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
18758 -maltivec=be support.
18759 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
18760 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
18761 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
18762 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
18763 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
18764 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
18765 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
18766 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
18767 altivec_vsumsws): Adjust.
18768 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
18769 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
18770 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
18772 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
18773 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
18774 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
18775 (altivec_lve<VI_char>x): Delete expand.
18776 (*altivec_lve<VI_char>x_internal): Rename to...
18777 (altivec_lve<VI_char>x): ... this.
18778 (altivec_lvxl_<mode>): Delete expand.
18779 (*altivec_lvxl_<mode>_internal): Rename to ...
18780 (altivec_lvxl_<mode>): ... this.
18781 (altivec_stvxl_<mode>): Delete expand.
18782 (*altivec_stvxl_<mode>_internal): Rename to ...
18783 (altivec_stvxl_<mode>): ... this.
18784 (altivec_stve<VI_char>x): Delete expand.
18785 (*altivec_stve<VI_char>x_internal): Rename to ...
18786 (altivec_stve<VI_char>x): ... this.
18787 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
18788 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
18789 reduc_plus_scal_<mode>): Adjust.
18790 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
18792 (rs6000_cpu_cpp_builtins): Adjust.
18793 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
18794 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
18795 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
18796 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
18797 -maltivec=be support.
18798 (rs6000_split_vec_extract_var): Adjust.
18799 (rs6000_split_v4si_init): Adjust.
18800 (swap_selector_for_mode): Delete.
18801 (altivec_expand_lvx_be, altivec_expand_stvx_be,
18802 altivec_expand_stvex_be): Delete.
18803 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
18804 -maltivec=be support.
18805 (rs6000_gimple_fold_builtin): Ditto.
18806 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
18808 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
18809 (TARGET_DIRECT_MOVE_64BIT): Adjust.
18810 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
18811 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
18812 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
18813 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
18814 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
18815 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
18816 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
18817 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
18818 anonymous split): Adjust.
18819 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
18820 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
18822 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
18824 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
18825 when --with-gxx-include-dir is also specified.
18826 * configure: Regenerate.
18828 2018-05-09 Jim Wilson <jimw@sifive.com>
18831 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
18832 * config/riscv/t-withmultilib: New.
18833 * config/riscv/withmultilib.h: New.
18834 * doc/install.texi: Document RISC-V --with-multilib-list support.
18836 2018-05-09 Richard Biener <rguenther@suse.de>
18838 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
18840 (vect_bb_vectorization_profitable_p): Adjust. Compute
18841 actual scalar cost using the cost vector and the add_stmt_cost
18844 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
18846 PR rtl-optimization/85645
18847 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
18848 in the REG_CFA_REGISTER note for LR, don't leave it empty.
18850 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
18852 PR rtl-optimization/85645
18853 * shrink-wrap.c (spread_components): Return a boolean saying if
18854 anything was changed.
18855 (try_shrink_wrapping_separate): Iterate spread_components until
18856 nothing changes anymore.
18858 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
18860 PR rtl-optimization/85645
18861 * regrename.c (build_def_use): Also kill the chains that include the
18862 destination of a REG_CFA_REGISTER note.
18864 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
18866 PR rtl-optimization/85645
18867 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
18868 insn that has a REG_CFA_REGISTER note.
18870 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
18872 * cfgexpand.c (expand_clobber): New function.
18873 (expand_gimple_stmt_1): Use it.
18874 * tree-vect-stmts.c (vect_clobber_variable): New function,
18876 (vectorizable_simd_clone_call): ...here.
18877 (vectorizable_store): Emit a clobber either side of an
18878 IFN_STORE_LANES sequence.
18879 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
18881 2018-05-09 Tom de Vries <tom@codesourcery.com>
18884 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
18885 (define_insn "trap_if_false"): Add exit after trap.
18887 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
18889 PR rtl-optimization/85638
18890 * bb-reorder.c: Include common/common-target.h.
18891 (create_forwarder_block): New function extracted from...
18892 (fix_up_crossing_landing_pad): ...here. Rename into...
18893 (dw2_fix_up_crossing_landing_pad): ...this.
18894 (sjlj_fix_up_crossing_landing_pad): New function.
18895 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
18896 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
18897 from both partitions and exit the loop after one iteration.
18899 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
18902 * doc/extend.texi (PowerPC Built-in Functions): Rename this
18904 (Basic PowerPC Built-in Functions): The new name of the
18905 subsection previously known as "PowerPC Built-in Functions".
18906 (Basic PowerPC Built-in Functions Available on all Configurations):
18908 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
18910 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
18912 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
18914 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
18917 2018-05-08 Jim Wilson <jimw@sifive.com>
18919 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
18920 (LD_EMUL_SUFFIX): New.
18921 (LINK_SPEC): Use it.
18923 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
18925 * doc/extend.texi (PowerPC Built-in Functions): Rename this
18927 (Basic PowerPC Built-in Functions): The new name of the
18928 subsection previously known as "PowerPC Built-in Functions".
18929 (Basic PowerPC Built-in Functions Available on all Configurations):
18931 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
18933 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
18935 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
18937 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
18940 2018-05-08 Jakub Jelinek <jakub@redhat.com>
18943 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
18944 after cmpelim optimization.
18946 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
18948 * config.gcc: Support "goldmont".
18949 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
18950 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18951 PROCESSOR_GOLDMONT.
18952 * config/i386/i386.c (m_GOLDMONT): Define.
18953 (processor_target_table): Add "goldmont".
18954 (PTA_GOLDMONT): Define.
18955 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
18956 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
18957 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
18958 (fold_builtin_cpu): Add "goldmont".
18959 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
18960 (ix86_option_override_internal): Add "goldmont".
18961 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
18962 (processor_type): Add PROCESSOR_GOLDMONT.
18963 * config/i386/i386.md: Add CPU "glm".
18964 * config/i386/glm.md: New file.
18965 * config/i386/x86-tune.def: Add m_GOLDMONT.
18966 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
18968 2018-05-08 Jakub Jelinek <jakub@redhat.com>
18971 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
18973 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
18974 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
18975 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
18978 * config/i386/i386.c (ix86_fold_builtin): Handle
18979 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
18982 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
18983 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
18985 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
18988 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
18989 (check_arch): Likewise.
18990 (check_fpu): Return the result rather than printing it.
18991 (end arch): Fix operator precedence.
18992 (end cpu): Likewise.
18993 (END): Print the result from check_fpu.
18995 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
18996 Alan Hayward <alan.hayward@arm.com>
18997 David Sherwood <david.sherwood@arm.com>
18999 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
19000 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
19001 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
19002 (*fcmuo<mode>_and): New patterns.
19004 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
19006 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
19007 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
19008 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
19009 (cmp_op, sve_imm_con): New code attributes.
19010 (SVE_COND_INT_CMP, imm_con): Delete.
19011 (cmp_op): Remove above unspecs from int attribute.
19012 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
19014 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
19015 comparison-specific unspecs.
19016 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
19017 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
19018 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
19019 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
19020 (*vec_fcm<cmp_op><mode>): Rename to...
19021 (*fcm<cmp_op><mode>): ...this and adjust likewise.
19022 (*vec_fcmuo<mode>): Rename to...
19023 (*fcmuo<mode>): ...this and adjust likewise.
19024 (*pred_fcm<cmp_op><mode>): New pattern.
19025 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
19026 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
19028 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
19030 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
19031 (aarch64_emit_sve_predicated_cond): New function.
19032 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
19033 (aarch64_emit_unspec_cond_or): Replace with...
19034 (aarch64_emit_sve_or_conds): ...this new function. Use
19035 aarch64_emit_sve_ptrue_op for the individual comparisons and
19036 aarch64_emit_binop to OR them together.
19037 (aarch64_emit_inverted_unspec_cond): Replace with...
19038 (aarch64_emit_sve_inverted_cond): ...this new function. Use
19039 aarch64_emit_sve_ptrue_op for the comparison and
19040 aarch64_emit_unop to invert the result.
19041 (aarch64_expand_sve_vec_cmp_float): Update after the above
19042 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
19044 2018-05-07 Nathan Sidwell <nathan@acm.org>
19046 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
19047 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
19048 (Backwards Compatibility): Likewise.
19050 2018-05-07 Luis Machado <luis.machado@linaro.org>
19054 2018-05-07 Luis Machado <luis.machado@linaro.org>
19056 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19057 <prefetch_dynamic_strides>: New const bool field.
19058 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19059 prefetch_dynamic_strides.
19060 (exynosm1_prefetch_tune): Likewise.
19061 (thunderxt88_prefetch_tune): Likewise.
19062 (thunderx_prefetch_tune): Likewise.
19063 (thunderx2t99_prefetch_tune): Likewise.
19064 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
19066 (aarch64_override_options_internal): Update to set
19067 PARAM_PREFETCH_DYNAMIC_STRIDES.
19068 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19069 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
19070 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
19071 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19072 prefetch-dynamic-strides setting.
19074 2018-05-07 Luis Machado <luis.machado@linaro.org>
19076 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19077 <minimum_stride>: New const int field.
19078 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19079 minimum_stride field.
19080 (exynosm1_prefetch_tune): Likewise.
19081 (thunderxt88_prefetch_tune): Likewise.
19082 (thunderx_prefetch_tune): Likewise.
19083 (thunderx2t99_prefetch_tune): Likewise.
19084 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
19085 (aarch64_override_options_internal): Update to set
19086 PARAM_PREFETCH_MINIMUM_STRIDE.
19087 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
19088 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
19089 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
19090 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
19091 stride is constant and is below the minimum stride threshold.
19093 2018-05-07 Luis Machado <luis.machado@linaro.org>
19095 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
19098 2018-05-07 Luis Machado <luis.machado@linaro.org>
19100 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19101 <prefetch_dynamic_strides>: New const bool field.
19102 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19103 prefetch_dynamic_strides.
19104 (exynosm1_prefetch_tune): Likewise.
19105 (thunderxt88_prefetch_tune): Likewise.
19106 (thunderx_prefetch_tune): Likewise.
19107 (thunderx2t99_prefetch_tune): Likewise.
19108 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
19110 (aarch64_override_options_internal): Update to set
19111 PARAM_PREFETCH_DYNAMIC_STRIDES.
19112 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19113 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
19114 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
19115 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19116 prefetch-dynamic-strides setting.
19118 2018-05-07 Luis Machado <luis.machado@linaro.org>
19120 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19121 <minimum_stride>: New const int field.
19122 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19123 minimum_stride field.
19124 (exynosm1_prefetch_tune): Likewise.
19125 (thunderxt88_prefetch_tune): Likewise.
19126 (thunderx_prefetch_tune): Likewise.
19127 (thunderx2t99_prefetch_tune): Likewise.
19128 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
19129 (aarch64_override_options_internal): Update to set
19130 PARAM_PREFETCH_MINIMUM_STRIDE.
19131 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
19132 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
19133 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
19134 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
19135 stride is constant and is below the minimum stride threshold.
19137 2018-05-06 Jakub Jelinek <jakub@redhat.com>
19140 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
19141 the type is addressable. Don't force op into register if it has
19144 2018-05-05 Roland McGrath <mcgrathr@google.com>
19147 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
19148 any section for which we don't know a specific type it should have,
19149 regardless of name. Previously this was done only for the exact
19150 names ".init_array", ".fini_array", and ".preinit_array".
19151 (default_elf_asm_named_section): Add comment about
19152 relationship with default_section_type_flags and SECTION_NOTYPE.
19153 (get_section): Don't consider it a type conflict if one side has
19154 SECTION_NOTYPE and the other doesn't, as long as neither has the
19155 SECTION_BSS et al used in the default_section_type_flags logic.
19157 2018-05-05 Tom de Vries <tom@codesourcery.com>
19160 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
19161 (workaround_barsyncs): New function.
19162 (nvptx_reorg): Use workaround_barsyncs.
19163 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
19164 (define_expand "nvptx_membar_cta"): New define_expand.
19165 (define_insn "*nvptx_membar_cta"): New insn.
19167 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
19169 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
19170 To improve optimization opportunities.
19171 * builtin-types.def: The new needed builtin types for the above.
19173 2018-05-04 Richard Biener <rguenther@suse.de>
19175 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
19176 * gimple-ssa-store-merging.c
19177 (imm_store_chain_info::output_merged_store): Remove redundant create,
19178 release split_store vector contents on failure.
19179 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
19180 scalar stmt vector on cache hit.
19182 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
19184 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
19186 * config.gcc (powerpc-xilinx-eabi*): Remove.
19187 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
19189 (fusion_addis_mem_combo_load): Ditto.
19190 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
19192 (rs6000_cpu_cpp_builtins): Ditto.
19193 * config/rs6000/rs6000-linux.c
19194 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
19195 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
19196 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
19198 (rs6000_setup_reg_addr_masks): Ditto.
19199 (rs6000_init_hard_regno_mode_ok): Ditto.
19200 (rs6000_option_override_internal): Ditto.
19201 (legitimate_lo_sum_address_p): Ditto.
19202 (rs6000_legitimize_address): Ditto.
19203 (rs6000_legitimize_reload_address): Ditto.
19204 (rs6000_legitimate_address_p): Ditto.
19205 (abi_v4_pass_in_fpr): Ditto.
19206 (setup_incoming_varargs): Ditto.
19207 (rs6000_gimplify_va_arg): Ditto.
19208 (rs6000_split_multireg_move): Ditto.
19209 (rs6000_savres_strategy): Ditto.
19210 (rs6000_emit_prologue_components): Ditto.
19211 (rs6000_emit_epilogue_components): Ditto.
19212 (rs6000_emit_prologue): Ditto.
19213 (rs6000_emit_epilogue): Ditto.
19214 (rs6000_elf_file_end): Ditto.
19215 (rs6000_function_value): Ditto.
19216 (rs6000_libcall_value): Ditto.
19217 * config/rs6000/rs6000.h: Ditto.
19218 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
19219 (TARGET_MINMAX): ... this. New.
19220 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
19221 * config/rs6000/rs6000.md: Remove Xilinx FP support.
19222 (*movsi_internal1_single): Delete.
19223 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
19224 mfpu=, mxilinx-fpu): Delete.
19225 * config/rs6000/singlefp.h: Delete.
19226 * config/rs6000/sysv4.h: Remove Xilinx FP support.
19227 * config/rs6000/t-rs6000: Ditto.
19228 * config/rs6000/t-xilinx: Delete.
19229 * config/rs6000/titan.md: Adjust for fp_type removal.
19230 * config/rs6000/vsx.md: Remove Xilinx FP support.
19231 (VStype_simple): Delete.
19232 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
19233 * config/rs6000/xfpu.h: Delete.
19234 * config/rs6000/xfpu.md: Delete.
19235 * config/rs6000/xilinx.h: Delete.
19236 * config/rs6000/xilinx.opt: Delete.
19237 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
19238 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
19240 2018-05-04 Tom de Vries <tom@codesourcery.com>
19243 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
19246 2018-05-04 Richard Biener <rguenther@suse.de>
19248 PR middle-end/85627
19249 * tree-complex.c (update_complex_assignment): We are always in SSA form.
19250 (expand_complex_div_wide): Likewise.
19251 (expand_complex_operations_1): Likewise.
19252 (expand_complex_libcall): Preserve EH info of the original stmt.
19253 (tree_lower_complex): Handle removed blocks.
19254 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
19255 on complex multiplication and division libcall builtins.
19257 2018-05-04 Richard Biener <rguenther@suse.de>
19259 PR middle-end/85574
19260 * fold-const.c (negate_expr_p): Restrict negation of operand
19261 zero of a division to when we know that can happen without
19263 (fold_negate_expr_1): Likewise.
19265 2018-05-04 Jakub Jelinek <jakub@redhat.com>
19268 * real.h (real_nextafter): Declare.
19269 * real.c (real_nextafter): New function.
19270 * fold-const-call.c (fold_const_nextafter): New function.
19271 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
19272 CASE_CFN_NEXTTOWARD.
19273 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
19274 even when arg1_mode is different from arg0_mode.
19276 2018-05-03 Nathan Sidwell <nathan@acm.org>
19278 * doc/extend.texi (Deprecated Features): Remove
19279 -ffriend-injection.
19280 (Backwards Compatibility): Likewise.
19281 * doc/invoke.texi (C++ Language Options): Likewise.
19282 (C++ Dialect Options): Likewise.
19284 2018-05-03 Jakub Jelinek <jakub@redhat.com>
19287 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
19288 _mm512_mask_mullox_epi64): New intrinsics.
19290 2018-05-03 Tom de Vries <tom@codesourcery.com>
19293 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
19294 dump files): Add offload-tree.
19296 2018-05-03 Richard Biener <rguenther@suse.de>
19298 PR tree-optimization/85615
19299 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
19300 to loops not nested in BBs loop father to avoid creating multi-entry
19303 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19305 PR tree-optimization/70291
19306 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
19307 arguments. Change return type to tree. Emit libcall as a new
19308 statement rather than replacing existing one when inplace_p is true.
19309 (expand_complex_multiplication_components): New function.
19310 (expand_complex_multiplication): Expand floating-point complex
19311 multiplication using the above.
19312 (expand_complex_division): Rename inner_type parameter to type.
19313 Update expand_complex_libcall call-site.
19314 (expand_complex_operations_1): Update expand_complex_multiplication
19315 and expand_complex_division call-sites.
19317 2018-05-02 Jakub Jelinek <jakub@redhat.com>
19320 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
19321 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
19322 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
19323 the highest significant bit of the shift count mask is clear. In
19324 check whether and[sq]i3 is needed verify that all significant bits
19325 of the shift count other than the highest are set.
19327 2018-05-02 Tom de Vries <tom@codesourcery.com>
19330 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
19331 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
19332 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
19333 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
19334 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
19335 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
19336 __builtin_goacc_parlevel_size.
19338 2018-05-02 Richard Biener <rguenther@suse.de>
19340 PR tree-optimization/85597
19341 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
19342 do not use split vect_get_vec_defs call but call vect_get_slp_defs
19345 2018-05-02 Tom de Vries <tom@codesourcery.com>
19348 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
19349 dump files): Add ltrans-tree.
19351 2018-05-02 Tom de Vries <tom@codesourcery.com>
19354 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
19355 dump files): Add wpa-ipa.
19357 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
19359 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
19360 powerpc*-*-linux*paired* target.
19361 * config/rs6000/750cl.h: Delete.
19362 * config/rs6000/paired.h: Delete.
19363 * config/rs6000/paired.md: Delete.
19364 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
19366 * config/rs6000/rs6000-builtin.def: Remove paired float support.
19367 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
19368 comment. Remove paired float support.
19369 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
19370 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
19372 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
19373 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
19375 * config/rs6000/rs6000.c: Remove paired float support.
19376 (paired_expand_vector_init, paired_expand_vector_move,
19377 paired_emit_vector_compare, paired_emit_vector_cond_expr,
19378 (paired_expand_lv_builtin, paired_expand_stv_builtin,
19379 paired_expand_builtin, paired_expand_predicate_builtin,
19380 paired_init_builtins): Delete.
19381 * config/rs6000/rs6000.h: Remove paired float support.
19382 * config/rs6000/rs6000.md: Remove paired float support.
19383 (move_from_CR_ov_bit): Delete.
19384 * config/rs6000/rs6000.opt (mpaired): Delete.
19385 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
19386 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
19388 2018-05-02 Richard Biener <rguenther@suse.de>
19390 PR middle-end/85567
19391 * gimplify.c (gimplify_save_expr): When in SSA form allow
19392 SAVE_EXPRs to compute to SSA vars.
19394 2018-05-02 Jakub Jelinek <jakub@redhat.com>
19397 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
19398 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
19399 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
19400 clobber operands[2], instead use a new pseudo. Formatting fixes.
19402 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
19404 PR tree-optimization/85586
19405 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
19406 exit early for statements in the same group if the accesses are
19409 2018-05-02 Tom de Vries <tom@codesourcery.com>
19412 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
19415 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
19417 PR tree-optimization/85143
19418 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
19420 2018-05-01 Tom de Vries <tom@codesourcery.com>
19423 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
19424 not found" error message.
19426 2018-05-01 Tom de Vries <tom@codesourcery.com>
19429 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
19430 * vec.c (test_ordered_remove_if): New function.
19431 (vec_c_tests): Call test_ordered_remove_if.
19432 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
19433 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
19434 * tree-vect-patterns.c (vect_pattern_recog_1): Use
19435 VEC_ORDERED_REMOVE_IF.
19437 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19439 PR tree-optimization/82665
19440 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
19441 pointer subtraction where arguments come from a memchr call.
19443 2018-05-01 Jakub Jelinek <jakub@redhat.com>
19445 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
19446 --push-state --as-needed and --pop-state instead of --as-needed and
19447 --no-as-needed if ld supports it.
19448 * configure: Regenerated.
19451 * doc/install.texi2html: Replace _002d with - and _002a with * in
19452 generated html files using sed.
19454 2018-04-30 David Malcolm <dmalcolm@redhat.com>
19457 * gcc-rich-location.c (blank_line_before_p): New function.
19458 (use_new_line): New function.
19459 (gcc_rich_location::add_fixit_insert_formatted): New function.
19460 * gcc-rich-location.h
19461 (gcc_rich_location::add_fixit_insert_formatted): New function.
19463 2018-04-30 David Malcolm <dmalcolm@redhat.com>
19465 * selftest.c (assert_streq): Rename "expected" and "actual" to
19466 "val1" and "val2". Extend NULL-handling to cover both inputs
19467 symmetrically, while still requiring both to be non-NULL for a pass.
19468 * selftest.h (assert_streq): Rename "expected" and "actual" to
19470 (ASSERT_EQ): Likewise.
19471 (ASSERT_EQ_AT): Likewise.
19472 (ASSERT_KNOWN_EQ): Likewise.
19473 (ASSERT_KNOWN_EQ_AT): Likewise.
19474 (ASSERT_NE): Likewise.
19475 (ASSERT_MAYBE_NE): Likewise.
19476 (ASSERT_MAYBE_NE_AT): Likewise.
19477 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
19478 the assertion to pass.
19479 (ASSERT_STREQ_AT): Likewise.
19481 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
19483 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
19484 interaction with -pie.
19486 2018-04-30 David Malcolm <dmalcolm@redhat.com>
19488 * selftest.h: Fix alphabetization of per-source-file selftest
19491 2018-04-30 Jason Merrill <jason@redhat.com>
19493 PR c++/61982 - dead stores to destroyed objects.
19494 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
19497 2018-04-30 Jason Merrill <jason@redhat.com>
19499 * tree.c (build_clobber): New.
19500 * tree.h: Declare it.
19501 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
19503 2018-04-30 David Malcolm <dmalcolm@redhat.com>
19505 * diagnostic-show-locus.c (layout::layout): Update for
19506 location_get_source_line returning a char_span.
19507 (struct char_span): Move to input.h.
19508 (struct correction): Update for fields in char_span becoming
19510 (struct source_line): Update for location_get_source_line
19511 returning a char_span.
19512 (layout::print_line): Likewise.
19513 * edit-context.c (edited_file::print_content): Likewise.
19514 (edited_file::print_diff_hunk): Likewise.
19515 (edited_file::print_run_of_changed_lines): Likewise.
19516 (edited_file::get_num_lines): Likewise.
19517 (edited_line::edited_line): Likewise.
19518 * final.c (asm_show_source): Likewise.
19519 * input.c (location_get_source_line): Convert return type
19520 from const char * to char_span, losing the final "line_len"
19522 (dump_location_info): Update for the above.
19523 (get_substring_ranges_for_loc): Likewise. Use a char_span
19524 when handling the literal within the line.
19525 (test_reading_source_line): Update for location_get_source_line
19526 returning a char_span.
19527 * input.h (class char_span): Move here from
19528 diagnostic-show-locus.c, converting from a struct to a class.
19529 Make data members private.
19530 (char_span::operator bool): New.
19531 (char_span::length): New.
19532 (char_span::get_buffer): New.
19533 (char_span::operator[]): New.
19534 (char_span::subspan): Make const.
19535 (char_span::xstrdup): New.
19536 (location_get_source_line): Convert return type from const char *
19537 to char_span, losing the final "line_size" param.
19539 2018-04-30 Jan Hubicka <jh@suse.cz>
19541 * lto-wrapper.c (ltrans_priorities): New static var.
19542 (cmp_priority): New.
19543 (run_gcc): Read priorities and if doing parallel build order
19544 the Makefile by them.
19546 2018-04-30 David Malcolm <dmalcolm@redhat.com>
19548 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
19550 2018-04-30 Richard Biener <rguenther@suse.de>
19552 * tree-cfg.c (verify_address): Remove base argument, add
19553 flag whether to check TREE_ADDRESSABLE and do that.
19554 (verify_expr): Remove.
19555 (verify_types_in_gimple_reference): Add pieces from verify_expr.
19556 (verify_gimple_assign_single): Likewise.
19557 (verify_gimple_switch): Likewise.
19558 (verify_expr_location_1): Dereference tp once. Add (disabled)
19559 piece from verify_expr.
19560 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
19562 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
19564 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
19566 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
19568 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
19569 (small_data_pattern): Likewise.
19570 (arc_rewrite_small_data): Likewise.
19571 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
19572 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
19573 (get_symbol_alignment): New function.
19574 (legitimate_small_data_address_p): Likewise.
19575 (legitimate_scaled_address): Update, call
19576 legitimate_small_data_address_p.
19577 (output_sdata): New static variable.
19578 (arc_print_operand): Update how we handle small data operands.
19579 (arc_print_operand_address): Likewise.
19580 (arc_legitimate_address_p): Update, use
19581 legitimate_small_data_address_p.
19582 (arc_rewrite_small_data_p): Remove.
19583 (arc_rewrite_small_data_1): Likewise.
19584 (arc_rewrite_small_data): Likewise.
19585 (small_data_pattern): Likewise.
19586 (compact_sda_memory_operand): Update to use
19587 legitimate_small_data_address_p and get_symbol_alignment.
19588 (prepare_move_operands): Don't rewite sdata pattern.
19589 (prepare_extend_operands): Remove.
19590 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
19592 (zero_extendqisi2): Likewise.
19593 (zero_extendhisi2): Likewise.
19594 (extendqihi2): Likewise.
19595 (extendqisi2): Likewise.
19596 (extendhisi2): Likewise.
19597 (addsi3): Likewise.
19598 (subsi3): Likewise.
19599 (andsi3): Likewise.
19600 * config/arc/constraints.md (Usd): Change it to memory constraint.
19602 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
19604 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
19605 as source of std instructions.
19606 * config/arc/arc.md (movsi_insn): Update pattern predicate to
19607 allow 6-bit constants as source for store instructions.
19608 (movdi_insn): Update instruction pattern to allow 6-bit constants
19609 as source for store instructions.
19611 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
19613 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
19615 2018-04-30 Nathan Sidwell <nathan@acm.org>
19616 Sandra Loosemore <sandra@codesourcery.com>
19618 * dumpfile.c (dump_open): Allow '-' for stdout.
19619 * doc/invoke.texi (Developer Options): Document dump filename
19620 determination early. Document stdin/stdout selection.
19622 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
19624 Microblaze Target: PIC data text relative
19626 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
19627 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
19629 * config/microblaze/microblaze.h (microblaze_constant_address_p):
19630 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
19631 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
19632 New addressing mode for data-text relative position indepenedent code.
19633 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
19634 'ADDRESS_SYMBOLIC_TXT_REL'.
19635 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
19636 (microblaze_legitimate_pic_operand): Exclude function calls from
19637 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
19638 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
19640 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
19641 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
19642 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
19643 for 'address + offset'.
19644 (microblaze_expand_prologue): Add new function prologue call for
19646 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
19647 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
19648 table in case of TARGET_PIC_DATA_TEXT_REL.
19649 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
19650 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
19651 Add new macros 'UNSPEC_TEXT',
19652 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
19653 + exclude function calls from 'UNSPEC_PLT' in case of data text
19655 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
19656 new target hook for generating address diff vector tables in case of
19658 * doc/tm.texi : Regenerate.
19659 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
19660 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
19661 of addr diff vector generation.
19662 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
19663 target hook definition.
19664 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
19665 Add default function for generate_pic_addr_diff_vec -> flag_pic.
19666 * doc/invoke.texi (Add new pic option): Add new microblaze pic
19667 option for data text relative.
19669 2018-04-30 Richard Biener <rguenther@suse.de>
19671 * tree-chrec.h (evolution_function_is_constant_p): Remove
19673 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
19675 2018-04-30 Richard Biener <rguenther@suse.de>
19678 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
19680 2018-04-30 Richard Biener <rguenther@suse.de>
19682 PR tree-optimization/28364
19683 PR tree-optimization/85275
19684 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
19685 copying first exit test.
19687 2018-04-28 Mark Wielaard <mark@klomp.org>
19689 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
19690 dwarf_version >= 5.
19691 (dwarf_AT): Handle DW_AT_addr_base.
19692 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
19694 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
19697 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
19698 (*ashl<dwi>3_doubleword_mask_1): Ditto.
19699 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
19700 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
19702 2018-04-28 Richard Biener <rguenther@suse.de>
19704 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
19705 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
19706 to reflect use. Only add interesting stmts.
19708 2018-04-27 Martin Jambor <mjambor@suse.cz>
19711 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
19712 the jump function allows for passing through aggregate values.
19714 2018-04-27 David Malcolm <dmalcolm@redhat.com>
19716 * input.h (in_system_header_at): Convert from macro to inline
19718 (from_macro_expansion_at): Likewise.
19719 (from_macro_definition_at): Likewise.
19721 2018-04-27 Jeff Law <law@redhat.com>
19723 * config.gcc: Mark tile* targets as deprecated/obsolete.
19725 2018-04-27 Richard Biener <rguenther@suse.de>
19727 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
19730 2018-04-27 Richard Biener <rguenther@suse.de>
19732 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
19734 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
19736 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
19737 with Yd constraint. Set "preferred_for_speed" attribute from
19738 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
19739 with Yd constraint.
19740 (*movdi_internal): Ditto.
19741 (movti_interunit splitters): Remove
19742 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
19743 (movdi_interunit splitters): Ditto.
19744 * config/i386/constraints.md (Ye): Remove.
19745 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
19747 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19750 * config/aarch64/constraints.md (Usg): Limit to 31.
19751 (Usj): Limit to 63.
19753 2018-04-27 Jakub Jelinek <jakub@redhat.com>
19755 PR tree-optimization/85529
19756 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
19757 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
19758 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
19759 zero extension or masking of the MSB bit.
19760 (optimize_range_tests): Add FIRST_BB argument, pass it through
19761 to optimize_range_tests_var_bound.
19762 (maybe_optimize_range_tests, reassociate_bb): Adjust
19763 optimize_range_tests callers.
19765 2018-04-26 Richard Biener <rguenther@suse.de>
19766 Jakub Jelinek <jakub@redhat.com>
19768 * cgraph.h (symbol_table): Just declare debug method here.
19769 * symtab.c (symbol_table::debug): Define.
19771 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
19773 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
19775 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
19777 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
19778 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
19779 (*movdi_internal): Substitute Yi and Yj constraint with x
19780 and Ym and Yn constraint with y constraint. Update "isa"
19781 attribute and set "preferred_for_speed" attribute from
19782 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
19783 (*movsi_internal): Ditto.
19784 (*movdf_internal): Ditto.
19785 (*movsf_internal): Ditto.
19786 (*zero_extendsidi2): Ditto.
19787 * config/i386/sse.md (vec_set<mode>_0): Ditto.
19788 (sse2_loadld): Ditto.
19789 (*vec_extract<ssevecmodelower>_0): Ditto.
19790 (*vec_extractv4si_0_zext_sse4): Ditto.
19791 (vec_concatv2di): Ditto.
19792 (*vec_dup<mode>): Ditto.
19793 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
19794 * config/i386/constraints.md (Yi): Remove.
19799 2018-04-26 Nathan Sidwell <nathan@acm.org>
19801 * dumpfile.c (dump_open): New.
19802 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
19803 (dump_finish): Detect stdio/stderr by value not name.
19805 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
19807 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
19809 2018-04-26 Tom de Vries <tom@codesourcery.com>
19812 * config/nvptx/nvptx.c (verify_neutering_jumps)
19813 (verify_neutering_labels): New function
19814 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
19816 2018-04-26 Tom de Vries <tom@codesourcery.com>
19819 * config/nvptx/nvptx.c (needs_neutering_p): New function.
19820 (nvptx_single): Use needs_neutering_p to skip over insns that do not
19823 2018-04-26 Richard Biener <rguenther@suse.de>
19824 Tom de Vries <tom@codesourcery.com>
19827 * lto-streamer-out.c (output_function): Fixup loops if required to match
19828 discovery done in the reader.
19830 2018-04-26 Richard Biener <rguenther@suse.de>
19832 PR tree-optimization/85116
19833 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
19834 have a loop exit from the single latch predecessor. Remove
19835 case of header with just condition.
19836 (ch_base::copy_headers): Exclude infinite loops from any
19838 (pass_ch::execute): Record exits.
19840 2018-04-26 Richard Biener <rguenther@suse.de>
19842 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
19843 prologue cost vector and pass it to vect_get_load_cost.
19844 (vect_get_peeling_costs_all_drs): Likewise.
19845 (vect_peeling_hash_get_lowest_cost): Likewise.
19846 (vect_enhance_data_refs_alignment): Likewise.
19848 2018-04-26 Richard Biener <rguenther@suse.de>
19850 PR middle-end/85450
19851 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
19852 checking of integer<->pointer conversions.
19853 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
19854 sign-/zero-extending pointer types.
19855 (expand_omp_for_static_chunk): Likewise.
19857 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
19858 Jean Lee <xiaoyur347@gmail.com>
19860 * config/mips/mips.c (mips_asan_shadow_offset): New function.
19861 (TARGET_ASAN_SHADOW_OFFSET): Define.
19862 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
19863 true for -fsanitize=address.
19865 2018-04-25 Mark Wielaard <mark@klomp.org>
19867 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
19870 2018-04-25 Jakub Jelinek <jakub@redhat.com>
19872 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
19873 than "alu", remove explicit "memory" and "imm_disp" attributes.
19874 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
19876 PR middle-end/85414
19877 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
19878 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
19879 gen_lowpart_no_emit.
19881 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
19884 * config/i386/i386.c (ix86_expand_builtin): Change memory
19885 operand to XI, extend p0 to Pmode.
19886 * config/i386/i386.md: Change unspec volatile and operand
19887 1 mode to XI, change operand 0 mode to P.
19889 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
19891 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
19892 GET_MODE_MASK before any checking.
19893 (nds32_can_use_bset_p): Likewise.
19894 (nds32_can_use_btgl_p): Likewise.
19896 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
19898 * config/nds32/nds32-doubleword.md: New define_split pattern for
19899 illegal register number.
19901 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
19903 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
19905 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
19907 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
19909 2018-04-25 Richard Biener <rguenther@suse.de>
19911 * lto-streamer.h (LTO_major_version): Bump to 8.
19913 2018-04-25 Jakub Jelinek <jakub@redhat.com>
19915 * BASE-VER: Set to 9.0.0.
19917 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
19919 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
19920 in __abskf2 and __powikf2.
19922 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19925 * config/aarch64/constraints.md (Usg, Usj): New constraints.
19926 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
19927 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
19928 Use the above on operand 2. Reindent.
19929 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
19931 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
19934 * common/config/i386/i386-common.c (ix86_handle_option): Don't
19936 * config/i386/i386.opt (mcet): Removed.
19937 * doc/install.texi: Remove -mcet documentation.
19938 * doc/invoke.texi: Likewise.
19940 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
19943 * doc/install.texi: Remove -mcet from bootstrap-cet.
19945 2018-04-24 Jakub Jelinek <jakub@redhat.com>
19948 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
19949 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
19953 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
19954 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
19955 containing a CONST_VECTOR.
19957 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
19959 * doc/install.texi: Update newlib dependency for nvptx.
19961 2018-04-24 Jakub Jelinek <jakub@redhat.com>
19964 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
19965 instead of INTVAL when shifting x left.
19967 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
19969 PR tree-optimization/85478
19970 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
19971 vect_grouped_store_supported for single element vectors.
19973 2018-04-24 Richard Biener <rguenther@suse.de>
19976 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
19977 load cost increase to the case of non-constant step.
19979 2018-04-24 Jakub Jelinek <jakub@redhat.com>
19982 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
19983 destination if any_malformed_asm.
19985 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
19987 PR middle-end/85496
19988 * expr.c (store_field): In the bitfield case, if the value comes from
19989 a function call and is returned in registers by means of a PARALLEL,
19990 do not change the mode of the temporary unless BLKmode and VOIDmode.
19992 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
19994 PR rtl-optimization/85423
19995 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
19996 dependencies to debug insns when the previous insn is non-debug.
19998 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
20000 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
20001 enums into a single definition.
20002 (fls): Fix predicates and printing.
20005 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
20007 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
20008 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
20009 and short u6 immediate.
20010 (check_if_valid_sleep_operand): Remove.
20011 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
20013 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
20015 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
20016 flag_always_save_lp condition.
20017 * config/nds32/nds32.opt (malways-save-lp): New option.
20019 2018-04-22 Shiva Chen <shiva0217@gmail.com>
20021 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
20022 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
20023 * config/nds32/nds32.h
20024 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
20025 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
20027 2018-04-22 Shiva Chen <shiva0217@gmail.com>
20029 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
20030 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
20032 2018-04-22 Shiva Chen <shiva0217@gmail.com>
20033 Chung-Ju Wu <jasonwucj@gmail.com>
20035 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
20037 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
20038 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
20040 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
20042 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
20044 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
20046 * config/nds32/nds32-protos.h (nds32_data_alignment,
20047 nds32_local_alignment): Declare.
20048 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
20049 nds32_local_alignment): New functions.
20050 (TARGET_CONSTANT_ALIGNMENT): Define.
20051 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
20053 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
20055 * config/nds32/nds32.c
20056 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
20057 (TARGET_MODES_TIEABLE_P): Likewise.
20059 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
20061 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
20062 level Ofast and Og.
20064 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
20065 Chung-Ju Wu <jasonwucj@gmail.com>
20067 * config/nds32/constants.md (unspec_volatile_element): Add enum values
20068 for unaligned access.
20069 * config/nds32/nds32-intrinsic.c: Implementation of expanding
20071 * config/nds32/nds32-intrinsic.md: Likewise.
20072 * config/nds32/nds32_intrinsic.h: Likewise.
20073 * config/nds32/nds32.h (nds32_builtins): Likewise.
20074 * config/nds32/nds32.opt (munaligned-access): New option.
20075 * config/nds32/nds32.c (nds32_asm_file_start): Display
20076 flag_unaligned_access status.
20078 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
20080 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
20081 -mno-relax is present.
20082 * config/riscv/linux.h (LINK_SPEC): Ditto.
20084 2018-04-20 Martin Sebor <msebor@redhat.com>
20087 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
20089 (gimple_fold_builtin_stxcpy_chk): Same.
20090 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
20092 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
20095 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
20096 __powikf2 when long double is IEEE 128-bit.
20098 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
20100 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
20101 step to make sure stack always aligned.
20103 2018-04-20 Carl Love <cel@us.ibm.com>
20106 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
20107 size check for arg0.
20109 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
20110 Tom de Vries <tom@codesourcery.com>
20113 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
20114 Emit insns for calls too.
20115 (nvptx_find_par): Always look for worker-level predecessor insn.
20116 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
20118 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
20119 (nvptx_process_pars): Propagate frames for calls.
20121 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
20124 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
20126 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
20127 (ix86_handle_option): Don't handle OPT_mibt.
20128 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
20130 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
20132 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
20133 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
20134 (ix86_target_macros): Define __CET__ with flag_cf_protection
20135 for -fcf-protection.
20136 * config/i386/i386.c (isa2_opts): Remove -mibt.
20137 * config/i386/i386.h (TARGET_IBT): Removed.
20138 (TARGET_IBT_P): Likewise.
20139 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
20140 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
20141 * config/i386/i386.opt (mcet): Update help message.
20142 (mshstk): Likewise.
20144 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
20145 -mcet as an alias for -mshstk.
20147 2018-04-20 Richard Biener <rguenther@suse.de>
20149 PR middle-end/85475
20150 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
20151 complexity by forcing a single use of the multiply operand.
20153 2018-04-20 Martin Jambor <mjambor@suse.cz>
20156 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
20157 recursion dependency to only apply to non-clones.
20159 2018-04-20 Martin Jambor <mjambor@suse.cz>
20162 * ipa-cp.c (create_specialized_node): Check that clones of
20163 self-recursive edges exist during IPA-CP.
20165 2018-04-19 Toon Moene <toon@moene.org>
20167 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
20170 2018-04-19 Jakub Jelinek <jakub@redhat.com>
20172 PR tree-optimization/85467
20173 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
20174 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
20175 VECTOR_CST element to type.
20177 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
20180 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
20181 * config/i386/i386.md (builtin_setjmp_setup): Removed.
20182 (builtin_longjmp): Likewise.
20183 (save_stack_nonlocal): New pattern.
20184 (restore_stack_nonlocal): Likewise.
20186 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
20189 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
20190 Replace ASM_OUTPUT_LABEL with fprintf.
20192 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
20195 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
20196 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
20197 * config/i386/i386-c.c (ix86_target_macros_internal): Also
20198 define __IBT__ and __SHSTK__ for -fcf-protection.
20199 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
20201 (ix86_trampoline_init): Likewise.
20202 (x86_output_mi_thunk): Likewise.
20203 (ix86_notrack_prefixed_insn_p): Likewise.
20204 (ix86_option_override_internal): Don't disallow -fcf-protection.
20205 * config/i386/i386.md (rdssp<mode>): Also enable for
20207 (incssp<mode>): Likewise.
20208 (nop_endbr): Likewise.
20209 * config/i386/i386.opt (mcet): Change help message to built-in
20212 (mshstk): Likewise.
20213 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
20214 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
20215 enable CET built-in functions.
20217 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
20219 * common/config/i386/i386-common.c
20220 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
20221 OPTION_MASK_ISA_MOVDIRI_UNSET,
20222 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
20223 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
20224 * config.gcc (movdirintrin.h): New header.
20225 * config/i386/cpuid.h (bit_MOVDIRI,
20226 bit_MOVDIR64B): New bits.
20227 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
20229 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
20230 (VOID, PVOID, PCVOID)): New function types.
20231 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
20232 __builtin_ia32_directstoreu_u64,
20233 __builtin_ia32_movdir64b): New builtins.
20234 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
20235 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
20237 (ix86_valid_target_attribute_inner_p): Ditto.
20238 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
20239 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
20240 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
20241 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
20242 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
20243 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
20244 (movdiri<mode>, movdir64b_<mode>): New.
20245 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
20246 * config/i386/immintrin.h: Include movdirintrin.h.
20247 * config/i386/movdirintrin.h: New file.
20248 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
20250 2018-04-19 Richard Biener <rguenther@suse.de>
20252 PR middle-end/85455
20253 * cfg.c (clear_bb_flags): When loop state says we have
20254 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
20256 2018-04-19 Richard Biener <rguenther@suse.de>
20258 PR tree-optimization/84737
20259 * tree-vect-data-refs.c (vect_copy_ref_info): New function
20260 copying restrict info.
20261 (vect_setup_realignment): Use it.
20262 * tree-vectorizer.h (vect_copy_ref_info): Declare.
20263 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
20264 the first DR to all generated stores.
20265 (vectorizable_load): Likewise for loads.
20267 2018-04-19 Jakub Jelinek <jakub@redhat.com>
20269 PR tree-optimization/85446
20270 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
20271 the integral and pointer types to have the same precision.
20273 * doc/install.texi: Document --disable-cet being the default and
20276 2018-04-18 Martin Liska <mliska@suse.cz>
20278 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
20281 2018-04-18 Martin Liska <mliska@suse.cz>
20284 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
20287 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
20288 arguments if they are comparable.
20290 2018-04-18 Martin Liska <mliska@suse.cz>
20293 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
20296 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
20299 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
20302 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
20303 ENDBR after calling __morestack.
20305 2018-04-18 David Malcolm <dmalcolm@redhat.com>
20308 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
20309 by using gcc_base_ver to generate a gcc_driver_version, and use
20310 it when generating GCC_DRIVER_NAME.
20311 * configure: Regenerate.
20313 2018-04-18 Jakub Jelinek <jakub@redhat.com>
20316 * config.gcc: Obsolete powerpc*-*-*spe*.
20318 2018-04-17 Jakub Jelinek <jakub@redhat.com>
20321 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
20322 (stabstr_D): Change type of unum from unsigned int to
20323 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
20326 2018-04-17 Jim Wilson <jimw@sifive.com>
20329 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
20330 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
20331 Set arg_pointer_offset after using pretend_args_size.
20333 2018-04-17 Jakub Jelinek <jakub@redhat.com>
20335 PR rtl-optimization/85431
20336 * dse.c (record_store): Ignore zero width stores.
20339 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
20340 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
20341 __builtin_stack_restore rather than after it.
20342 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
20343 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
20344 argument instead of virtual_dynamic_stack_rtx.
20346 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
20348 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
20350 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20351 Add note to error message to explain internal mapping of overloaded
20352 built-in function name to non-overloaded built-in function name.
20353 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
20356 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
20359 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
20360 where the inputs overlap with the output.
20362 2018-04-17 Jakub Jelinek <jakub@redhat.com>
20364 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
20365 (=v, v) alternative and explicit "memory" attribute.
20366 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
20367 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
20369 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
20370 "sselog1" type instead of "sselog".
20371 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
20372 "sselog". Remove explicit "memory" attribute.
20373 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
20374 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
20376 (vec_extract_hi_v32hi): Merge all alternatives into one, use
20377 "sselog1" type instead of "sselog". Remove explicit "memory"
20379 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
20380 use "sselog1" type instead of "sselog". Remove explicit "memory"
20382 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
20383 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
20385 (vec_extract_hi_v64qi): Merge all alternatives into one, use
20386 "sselog1" type instead of "sselog". Remove explicit "memory"
20388 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
20389 use "sselog1" type instead of "sselog". Remove explicit "memory"
20393 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
20395 PR middle-end/85414
20396 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
20399 2018-04-17 Martin Jambor <mjambor@suse.cz>
20402 * ipa-cp.c (create_specialized_node): Call
20403 expand_all_artificial_thunks if necessary.
20405 2018-04-17 Martin Liska <mliska@suse.cz>
20408 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
20409 in message, remote space in between '_G' and '('.
20411 2018-04-17 Jakub Jelinek <jakub@redhat.com>
20414 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
20415 avx512f_vmcmp<mode>3<round_saeonly_name>,
20416 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
20417 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
20418 avx512f_rndscale<mode><round_saeonly_name>,
20419 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
20420 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
20421 Use %<iptr>2 instead of %2 for -masm=intel.
20422 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
20423 avx512f_vcvttss2usi<round_saeonly_name>,
20424 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
20426 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
20427 avx512f_vcvttsd2usi<round_saeonly_name>,
20428 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
20429 Use %q1 instead of %1 for -masm=intel.
20430 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
20431 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
20432 of %3 for -masm=intel.
20433 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
20435 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
20437 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
20439 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
20440 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
20442 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
20444 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
20445 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
20446 %g1 and one with %0 and %1.
20447 (avx512er_vmrcp28<mode><round_saeonly_name>,
20448 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
20449 %1 for -masm=intel.
20450 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
20451 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
20452 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
20453 of %0 and %{%4%} for -masm=intel.
20454 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
20455 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
20456 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
20457 order of %0 and %{%5%}%{z%} for -masm=intel.
20459 2018-04-17 Jan Hubicka <jh@suse.cz>
20462 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
20464 2018-04-17 Martin Liska <mliska@suse.cz>
20467 * multiple_target.c (create_dispatcher_calls): Set apostrophes
20468 for target_clone error message. Make default implementation
20469 clone to be a local declaration.
20470 (separate_attrs): Add new argument and check for an empty
20472 (expand_target_clones): Handle it.
20473 (ipa_target_clone): Make redirection just for target_clones
20476 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
20477 Tom de Vries <tom@codesourcery.com>
20479 PR middle-end/84955
20480 * omp-expand.c (expand_oacc_for): Add dummy false branch for
20481 tiled basic blocks without omp continue statements.
20483 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
20486 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
20487 vec_extract expression as having side effects to make sure it gets
20490 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
20493 * config/i386/i386.c (get_builtin_code_for_version): Check
20496 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
20499 * config.gcc: Support "skylake".
20500 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20502 * config/i386/i386.c (m_SKYLAKE): Define.
20503 (processor_target_table): Add "skylake".
20504 (ix86_option_override_internal): Add "skylake".
20505 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
20506 PROCESSOR_CANNONLAKE.
20507 (get_builtin_code_for_version): Fix priority for
20508 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
20509 PROCESSOR_SKYLAKE-AVX512.
20510 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
20511 (processor_type): Add PROCESSOR_SKYLAKE.
20513 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
20514 Jason Merrill <jason@redhat.com>
20517 * convert.c (convert_to_integer_1): Use direct recursion for
20518 enumeral types and types with a precision less than the number
20519 of bits in their mode.
20521 2018-04-16 Julia Koval <julia.koval@intel.com>
20524 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
20525 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
20527 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
20530 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
20531 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
20532 and -mno-direct-move.
20534 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
20537 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
20538 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
20539 Ensure negative shifts result in {0}.
20541 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
20543 PR rtl-optimization/79916
20544 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
20545 regs (if any) to define how to gnerate SD moves when LRA is in
20548 2018-04-13 Jakub Jelinek <jakub@redhat.com>
20550 PR rtl-optimization/85393
20551 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
20552 * except.c (expand_dw2_landing_pad_for_region): Make static.
20553 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
20554 a label and unconditional jump to old_bb, rather than
20555 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
20558 PR rtl-optimization/85376
20559 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
20560 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
20561 instead of a specific value.
20563 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
20564 Bin Cheng <bin.cheng@arm.com>
20566 PR tree-optimization/82965
20567 PR tree-optimization/83991
20568 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
20569 by_profile_only parameter.
20570 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
20571 information if the loop was predicted to iterate too many times.
20572 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
20574 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
20577 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
20580 2018-04-13 Martin Liska <mliska@suse.cz>
20581 Jakub Jelinek <jakub@redhat.com>
20583 PR middle-end/81657
20584 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
20585 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
20586 * builtins.c (expand_builtin_memory_copy_args): Use
20587 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
20588 handle dest_addr == pc_rtx.
20590 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
20593 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
20594 asked to not generate direct moves.
20595 (fix_trunc<mode>si2_stfiwx): Similar.
20596 (fix_trunc<mode>si2_internal): Similar.
20598 2018-04-12 Jakub Jelinek <jakub@redhat.com>
20601 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
20602 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
20603 lookup if dest in some wider mode is known to be const0_rtx and
20604 if so, record permanent equivalence for it to be ZERO_EXTEND of
20605 the narrower mode destination.
20607 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
20609 * lto-streamer-out.c (output_function): Revert 259346.
20610 * omp-expand.c (expand_oacc_for): Likewise.
20612 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
20614 PR rtl-optimization/85354
20615 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
20616 * sel-sched.c (sel_global_init): ... here.
20618 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
20621 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
20622 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
20623 mode for PE-COFF targets.
20624 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
20625 (i386_pe_asm_lto_end): Likewise.
20626 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
20627 (TARGET_ASM_LTO_END): Likewise.
20628 * config/i386/winnt.c (saved_debug_info_level): New static variable.
20629 (i386_pe_asm_lto_start): New function.
20630 (i386_pe_asm_lto_end): Likewise.
20632 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
20633 Richard Biener <rguenther@suse.de>
20635 PR middle-end/84955
20636 * lto-streamer-out.c (output_function): Fix CFG loop state before
20638 * omp-expand.c (expand_oacc_for): Handle calls to internal
20639 functions like regular functions.
20641 2018-04-12 Richard Biener <rguenther@suse.de>
20644 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
20645 for the early LTO debug to properly generate references to it
20646 during DIE emission. Do not re-use that for the skeleton for
20648 (dwarf2out_early_finish): Likewise.
20650 2018-04-12 Jakub Jelinek <jakub@redhat.com>
20653 * config/i386/sse.md
20654 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
20655 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
20656 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
20657 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
20658 and output is a reg, avoid creating invalid lowpart subreg, but
20659 instead split into a 512-bit move. Don't split if not AVX512VL,
20660 input is xmm16+ reg and output is a mem.
20661 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
20662 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
20663 xmm16+ reg and output is a mem.
20665 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20667 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
20668 also for flag_dwarf2_cfi_asm.
20670 2018-04-12 Jakub Jelinek <jakub@redhat.com>
20672 PR rtl-optimization/85342
20673 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
20674 a bool scalar var inside of the loop instead. Don't try to update
20675 recog_data.operand after failed apply_change_group.
20677 2018-04-12 Tom de Vries <tom@codesourcery.com>
20680 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
20681 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
20682 array with flexible array member as array without given dimension.
20683 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
20684 argument for undefined param to true.
20686 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
20689 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
20690 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
20691 from PowerPC section.
20692 * config/rs6000/sysv4.opt (mcall-): Improve help text.
20693 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
20694 help text that is too long.
20695 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
20696 help text that is too long.
20697 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
20698 help text that is too long.
20700 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
20702 * config/alpha/alpha.md (stack_probe_internal): Rename
20703 from "probe_stack". Update all callers.
20705 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
20707 PR rtl-optimization/84566
20708 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
20709 sched_macro_fuse_insns.
20711 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
20714 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
20715 (compute_block_dependences): ... from here.
20717 2018-04-11 Jakub Jelinek <jakub@redhat.com>
20719 PR tree-optimization/85331
20720 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
20721 from int to HOST_WIDE_INT.
20723 2018-04-11 Martin Jambor <mjambor@suse.cz>
20726 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
20727 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
20728 not the same as the source val.
20729 (cgraph_edge_brings_value_p): New parameter.
20730 (gather_edges_for_value): Pass destination value to
20731 cgraph_edge_brings_value_p.
20732 (perhaps_add_new_callers): Likewise.
20733 (get_info_about_necessary_edges): Likewise and exclude values brought
20734 only by self-recursive edges.
20735 (create_specialized_node): Redirect only clones of self-calling edges.
20736 (+self_recursive_pass_through_p): New function.
20737 (find_more_scalar_values_for_callers_subset): Use it.
20738 (find_aggregate_values_for_callers_subset): Likewise.
20739 (known_aggs_to_agg_replacement_list): Removed.
20740 (decide_whether_version_node): Re-calculate known constants for all
20741 remaining context clones.
20743 2018-04-11 Richard Biener <rguenther@suse.de>
20746 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
20747 from early DWARF output.
20748 (dwarf2out_early_finish): Output line info unconditionally into
20749 early DWARF and add reference to it.
20751 2018-04-11 Jakub Jelinek <jakub@redhat.com>
20754 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
20755 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
20756 other than V2DFmode using iptr mode attribute.
20757 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
20759 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
20761 PR rtl-optimization/84659
20762 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
20764 2018-04-11 Jakub Jelinek <jakub@redhat.com>
20767 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
20769 (output_loc_list): Pass address of a dummy size variable even in the
20770 locview handling loop.
20771 (index_location_lists): Add comment on why skip_loc_list_entry can't
20774 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
20777 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
20780 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
20783 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
20784 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
20785 and -mstring-compare-inline-limit.
20787 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
20790 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
20791 for stack clash protection in a register whenever we need it to be in
20794 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
20796 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
20797 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
20799 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
20802 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
20804 (mlong-double-): Ditto.
20805 * config/rs6000/sysv4.opt (msdata=): Ditto.
20806 (mtls-size=): Ditto.
20808 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
20810 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
20811 erroneous entries for
20812 "vector int vec_ldl (int, long int *)", and
20813 "vector unsigned int vec_ldl (int, unsigned long int *)".
20814 Add comments and entries for
20815 "vector bool char vec_ldl (int, bool char *)",
20816 "vector bool short vec_ldl (int, bool short *)",
20817 "vector bool int vec_ldl (int, bool int *)",
20818 "vector bool long long vec_ldl (int, bool long long *)",
20819 "vector pixel vec_ldl (int, pixel *)",
20820 "vector long long vec_ldl (int, long long *)",
20821 "vector unsigned long long vec_ldl (int, unsigned long long *)".
20822 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
20823 type tree bool_long_long_type_node and correct definition of
20824 bool_V2DI_type_node to make reference to this new type tree.
20825 (rs6000_mangle_type): Replace erroneous reference to
20826 bool_long_type_node with bool_long_long_type_node.
20827 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
20828 comments to emphasize sign distinctions for char and int types and
20829 replace RS6000_BTI_bool_long constant with
20830 RS6000_BTI_bool_long_long constant. Also add comment to restrict
20831 use of RS6000_BTI_pixel.
20832 (bool_long_type_node): Remove this macro definition.
20833 (bool_long_long_type_node): New macro definition
20835 2018-04-10 Jakub Jelinek <jakub@redhat.com>
20837 PR rtl-optimization/85300
20838 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
20839 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
20840 simplify_unary_operation fails.
20842 2018-04-10 Martin Liska <mliska@suse.cz>
20844 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
20845 cgraph_edge and ipa_ref.
20847 2018-04-10 Jakub Jelinek <jakub@redhat.com>
20851 * config/i386/sse.md
20852 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
20853 computation of the VEC_MERGE selector from mask.
20854 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
20855 Fix decoding of the VEC_MERGE selector into mask.
20857 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
20859 PR tree-optimization/85286
20860 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
20862 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
20864 * final.c (final_1): Set insn_last_address as well as
20865 insn_current_address.
20867 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20870 * explow.c (emit_stack_probe): Call validize_mem on memory location
20871 before passing it to gen_probe_stack. Create address operand and
20872 legitimize it for the probe_stack_address case.
20874 2018-04-09 Jan Hubicka <jh@suse.cz>
20877 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
20878 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
20879 * tree.c (free_lang_data_in_type): Fix handling of binfos;
20881 (free_lang_data): Rebuild type inheritance graph.
20883 2018-04-09 Martin Sebor <msebor@redhat.com>
20885 * invoke.texi (-finline-small-functions): Mention other optimization
20887 (-findirect-inlining, -fpartial-inlining): Same.
20888 (-finline-functions-called-once): Same.
20889 (-freorder-blocks-and-partition): Same.
20891 2018-04-09 Jan Hubicka <jh@suse.cz>
20894 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
20895 jumps; choose last target that matches the criteria (i.e.
20896 no partition changes for non-crossing jumps).
20897 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
20898 support for redirecting crossing jumps to non-crossing.
20900 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
20902 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
20903 also for naked functions.
20905 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
20907 * config/arc/arc.md (add_shift): New pattern.
20908 (add_shift2): Likewise.
20909 (sub_shift): Likewise.
20910 (sub_shift_cmp0_noout): Likewise.
20911 (compare_si_ashiftsi): Likewise.
20912 (xbfu_cmp0_noout): New combine pattern.
20913 (xbfu_cmp0"): Likewise.
20914 (movsi_set_cc_insn): Place the predicable variant first.
20915 (commutative_binary_cmp0_noout): Remove clobber.
20916 (commutative_binary_cmp0): New pattern.
20917 (noncommutative_binary_cmp0): Likewise.
20918 (noncommutative_binary_cmp0_noout): Likewise.
20919 (noncommutative_binary_comparison_result_used): Removed.
20920 (rsub_cmp0): New pattern.
20921 (rsub_cmp0_noout): Likewise.
20922 (extzvsi): Changed, keep only meaningful variants.
20923 (SQH, SEZ): New iterators.
20924 (SQH_postfix): New mode attribute.
20925 (SEZ_prefix): New code attribute.
20926 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
20927 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
20928 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
20929 of numerical value.
20930 (noncommutative_operator): Check the availability of barrel
20933 2018-04-09 Richard Biener <rguenther@suse.de>
20935 PR tree-optimization/85284
20936 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
20937 Only use the niter constraining form of simple_iv when the exit
20938 is always executed.
20940 2018-04-09 Tom de Vries <tom@codesourcery.com>
20943 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
20944 (define_expand "*memory_barrier"): New define_expand.
20945 (define_insn "memory_barrier"): New insn.
20947 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
20949 PR rtl-optimization/80463
20950 PR rtl-optimization/83972
20951 PR rtl-optimization/83480
20953 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
20954 correct producer for the insn.
20955 (tidy_control_flow): Fixup seqnos in case of debug insns.
20957 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
20959 PR rtl-optimization/83913
20961 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
20962 different sched-times when merging exprs.
20964 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
20966 PR rtl-optimization/83962
20968 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
20969 tidy_fallthru_edge and tidy_control_flow.
20971 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
20973 PR rtl-optimization/83530
20975 * sel-sched.c (force_next_insn): New global variable.
20976 (remove_insn_for_debug): When force_next_insn is true, also leave only
20977 next insn in the ready list.
20978 (sel_sched_region): When the region wasn't scheduled, make another pass
20979 over it with force_next_insn set to 1.
20981 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
20983 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
20985 * config/nds32/constants.md (unspec_volatile_element): Add enum values
20986 for interrupt control.
20987 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
20988 functions for interrupt control.
20989 * config/nds32/nds32-intrinsic.md: Likewise.
20990 * config/nds32/nds32_intrinsic.h: Likewise.
20991 * config/nds32/nds32.h (nds32_builtins): Likewise.
20993 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
20995 * config/nds32/nds32.c (nds32_init_machine_status,
20996 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
20997 strict_aligned_p field.
20998 (nds32_expand_to_rtl_hook): New function.
20999 (TARGET_EXPAND_TO_RTL_HOOK): Define.
21000 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
21002 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
21003 Chung-Ju Wu <jasonwucj@gmail.com>
21005 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
21006 * config/nds32/nds32-n7.md: New file.
21007 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
21008 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
21010 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
21011 * config/nds32/nds32.md (pipeline_model): Add n7.
21012 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
21013 * config/nds32/pipelines.md: Include n7 settings.
21015 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
21016 Chung-Ju Wu <jasonwucj@gmail.com>
21018 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
21019 * config/nds32/nds32-e8.md: New file.
21020 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
21021 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
21023 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
21024 * config/nds32/nds32.md (pipeline_model): Add e8.
21025 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
21026 * config/nds32/pipelines.md: Include e8 settings.
21028 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
21029 Chung-Ju Wu <jasonwucj@gmail.com>
21031 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
21032 * config/nds32/nds32-n8.md: New file.
21033 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
21034 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
21036 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
21037 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
21038 * config/nds32/nds32.md (pipeline_model): Add n8.
21039 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
21040 * config/nds32/pipelines.md: Include n8 settings.
21042 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
21043 Chung-Ju Wu <jasonwucj@gmail.com>
21045 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
21046 * config/nds32/nds32-n9-2r1w.md: New file.
21047 * config/nds32/nds32-n9-3r2w.md: New file.
21048 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
21049 nds32_register_ports): New or modify for cpu n9.
21050 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
21052 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
21053 * config/nds32/nds32-utils.c: New file.
21054 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
21055 TARGET_MUL_SLOW): Define.
21056 * config/nds32/nds32.md (pipeline_model): New attribute.
21057 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
21058 New options that support cpu n9.
21059 * config/nds32/pipelines.md: Include n9 settings.
21060 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
21062 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
21064 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
21065 information if necessary.
21066 (output_cond_branch_compare_zero): Likewise.
21067 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
21068 (nds32_target_alignment): Refine for alignment.
21069 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
21070 (FUNCTION_BOUNDARY): Modify.
21071 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
21073 * config/nds32/nds32.opt (malways-align, malign-functions): New.
21075 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
21077 * config/nds32/constants.md (unspec_volatile_element): Add values for
21078 TLB operation and data prefetch.
21079 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
21080 functions for TLB operation and data prefetch.
21081 * config/nds32/nds32-intrinsic.md: Likewise.
21082 * config/nds32/nds32_intrinsic.h: Likewise.
21083 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
21084 (nds32_print_operand): Likewise.
21085 * config/nds32/nds32.h (nds32_builtins): Likewise.
21087 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
21088 Andrew Pinski <pinsika@gcc.gnu.org>
21090 PR middle-end/82976
21091 * match.pd: Use constant_boolean_node of correct type instead of
21092 boolean_true_node or boolean_false_node for simplifying
21093 pointer comparisons to zero.
21095 2018-04-07 Jakub Jelinek <jakub@redhat.com>
21097 PR tree-optimization/80021
21098 * tree.c (verify_type_variant): Make error call in verify_variant_match
21099 translatable and remove final full stop.
21101 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
21103 * config/nds32/constants.md (unspec_volatile_element): Add
21104 UNSPEC_VOLATILE_EH_RETURN.
21105 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
21106 nds32_output_stack_pop): Support dwarf exception handling process.
21107 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
21108 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
21109 exception handling process.
21110 (nds32_compute_stack_frame): Likewise.
21111 (nds32_return_addr_rtx): Likewise.
21112 (nds32_initial_elimination_offset): Likewise.
21113 (nds32_expand_prologue): Likewise.
21114 (nds32_expand_epilogue): Likewise.
21115 (nds32_dynamic_chain_address): New function.
21116 * config/nds32/nds32.h (machine_function): Add fields for dwarf
21117 exception handling.
21118 (DYNAMIC_CHAIN_ADDRESS): Define.
21119 (EH_RETURN_DATA_REGNO): Define.
21120 (EH_RETURN_STACKADJ_RTX): Define.
21121 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
21122 patterns for dwarf exception handling.
21124 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
21126 * config/nds32/nds32.h: Clean up obsolete macros.
21128 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
21130 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
21131 Add enum values for particular instructions.
21132 * config/nds32/nds32-intrinsic.c: Implementation of expanding
21133 particular intrinsic functions.
21134 * config/nds32/nds32-intrinsic.md: Likewise.
21135 * config/nds32/nds32_intrinsic.h: Likewise.
21136 * config/nds32/nds32.h (nds32_builtins): Likewise.
21137 * config/nds32/nds32.md (type): Add pbsad and pbsada.
21138 (btst, ave): New patterns for particular instructions.
21140 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
21142 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
21143 Add enum values for atomic load/store and memory sync.
21144 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
21146 * config/nds32/nds32-intrinsic.md: Likewise.
21147 * config/nds32/nds32_intrinsic.h: Likewise.
21148 * config/nds32/nds32.h (nds32_builtins): Likewise.
21150 2018-04-07 Jakub Jelinek <jakub@redhat.com>
21152 PR tree-optimization/85257
21153 * fold-const.c (native_encode_vector): If not all elts could fit
21154 and off is -1, return 0 rather than offset.
21155 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
21156 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
21157 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
21158 adjust buffer in native_interpret_expr call.
21160 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
21162 * config/nds32/constants.md (unspec_volatile_element): Add cache
21163 control enum values.
21164 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
21165 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
21166 * config/nds32/nds32.c (nds32_cctl_names): New.
21167 (nds32_print_operand): Handle cache control register names.
21168 * config/nds32/nds32.h (nds32_builtins): New enum values.
21169 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
21171 * config/nds32/nds32.md (type): Add mmu.
21172 * config/nds32/pipelines.md (simple_insn): Add mmu.
21174 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
21176 * config/nds32/nds32.md (type): Remove call.
21177 * config/nds32/pipelines.md (simple_insn): Likewise.
21179 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
21181 * config/nds32/constants.md (unspec_volatile_element): Add
21182 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
21183 UNSPEC_VOLATILE_FMFCFG.
21184 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
21185 description for fmfcfg and fmfcsr.
21186 (bdesc_1arg): Add fmtcsr.
21187 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
21188 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
21189 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
21190 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
21191 unspec_fmfcfg): New patterns.
21192 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
21193 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
21194 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
21195 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
21196 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
21197 __nds32__fmfcfg): Define.
21199 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
21201 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
21202 intrinsic register names.
21203 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
21204 intrinsic register enum values and macros.
21206 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
21208 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
21209 for load/store addressing form.
21210 (nds32_print_operand_address): Likewise.
21212 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
21215 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
21216 based on LABEL_REF. Remove useless assertion.
21217 (pic_address_needs_scratch): Fix formatting.
21218 (sparc_legitimize_pic_address): Minor tweaks.
21219 (sparc_delegitimize_address): Adjust assertion accordingly.
21220 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
21221 into symbolic_operand.
21222 (movsi_high_pic_label_ref): Likewise.
21223 (movsi_lo_sum_pic_label_ref): Likewise.
21224 (movdi_pic_label_ref): Likewise.
21225 (movdi_high_pic_label_ref): Likewise.
21226 (movdi_lo_sum_pic_label_ref): Likewise.
21228 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
21230 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
21231 custom LIB_SPEC setup.
21233 2018-04-06 Ruslan Bukin <br@bsdpad.com>
21234 Kito Cheng <kito.cheng@gmail.com>
21236 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
21237 * config/riscv/freebsd.h: New.
21239 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
21241 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
21242 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
21245 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
21246 Kito Cheng <kito.cheng@gmail.com>
21248 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
21249 nds32_output_call, nds32_symbol_binds_local_p): New functions.
21250 * config/nds32/nds32-protos.h (nds32_output_call,
21251 nds32_output_return): Declare.
21252 * config/nds32/nds32.md: Refine all the call and return patterns.
21254 2018-04-06 Jakub Jelinek <jakub@redhat.com>
21257 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
21258 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
21260 PR rtl-optimization/84872
21261 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
21262 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
21263 EDGE_CROSSING edge.
21265 2018-04-06 Tamar Christina <tamar.christina@arm.com>
21267 * expr.c (copy_blkmode_to_reg): Revert 254862.
21268 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
21270 2018-04-06 Richard Biener <rguenther@suse.de>
21272 PR middle-end/85244
21273 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
21274 after seeing a component reference with an adjacent field. Treat
21275 refs to arrays at struct end of external decls similar to
21276 refs to unconstrained commons.
21278 2018-04-06 Jakub Jelinek <jakub@redhat.com>
21281 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
21282 look through SAVE_EXPRs with non-side-effects argument. Adjust
21284 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
21287 2018-04-06 Richard Biener <rguenther@suse.de>
21289 PR middle-end/85180
21290 * alias.c (find_base_term): New wrapper around find_base_term
21291 unwinding CSELIB_VAL_PTR changes.
21292 (find_base_term): Do not restore CSELIB_VAL_PTR during the
21295 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21297 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
21299 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
21300 constant definitions.
21301 ("nop"): lr 0,0 -> nopr r0
21302 ("nop_lr0", "nop_lr1"): New insn definitions.
21304 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
21306 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
21307 NDS32_V3PUSH_AVAILABLE_P macro.
21309 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
21310 Chung-Ju Wu <jasonwucj@gmail.com>
21312 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
21313 (nds32*-*-*): Add float and fpu_config into supported_defaults.
21314 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
21315 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
21316 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
21317 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
21318 * config/nds32/constraints.md: New constraints and checking for hard
21319 float configuration.
21320 * config/nds32/iterators.md: New mode iterator and attribute for hard
21321 float configuration.
21322 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
21324 * config/nds32/nds32-fpu.md: New file.
21325 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
21326 deal with hard float code generation.
21327 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
21329 (abi_type, float_reg_number): New enum type.
21330 * config/nds32/nds32-predicates.c: New predicates for hard float.
21331 * config/nds32/nds32-protos.h: Declare functions for hard float.
21332 * config/nds32/nds32.c: Implementation for hard float configuration.
21333 * config/nds32/nds32.h: Definitions for hard float configuration.
21334 * config/nds32/nds32.md: Include hard float machine description and
21335 modify patterns for hard float configuration.
21336 * config/nds32/nds32.opt: New options for hard float configuration.
21337 * config/nds32/predicates.md: New predicates for hard float
21340 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
21342 * common/config/nds32/nds32-common.c
21343 (nds32_option_optimization_table): Enable -mreleax-hint by default.
21345 2018-04-05 Jakub Jelinek <jakub@redhat.com>
21347 PR middle-end/85195
21348 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
21349 CONSTRUCTOR_ELT (ctor, ...)->value.
21351 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
21354 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
21356 2018-04-05 Tom de Vries <tom@codesourcery.com>
21359 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
21362 2018-04-05 Shiva Chen <shiva0217@gmail.com>
21363 Kito Cheng <kito.cheng@gmail.com>
21365 * config/nds32/constraints.md (U33): Fine-tune checking condition.
21366 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
21367 * config/nds32/nds32.h (nds32_16bit_address_type): Add
21368 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
21370 2018-04-05 Shiva Chen <shiva0217@gmail.com>
21371 Kito Cheng <kito.cheng@gmail.com>
21373 * config/nds32/constraints.md (Ufe): New memory constraint.
21374 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
21375 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
21376 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
21378 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
21379 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
21381 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
21383 * config/nds32/nds32.md: Use optimize_size in the condition for
21384 alu-shift instructions.
21386 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
21388 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
21390 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
21392 * config/nds32/nds32.md (negsi2): Refine pattern.
21394 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
21395 Chung-Ju Wu <jasonwucj@gmail.com>
21397 * config/nds32/iterators.md (shift_rotate): New code iterator.
21398 (shift): New code attribute.
21399 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
21400 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
21401 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
21402 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
21403 bit-wise operations.
21404 (andsi3, *andsi3): Ditto.
21405 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
21406 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
21407 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
21408 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
21409 nds32_ior_operand, nds32_xor_operand): New predicates.
21411 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
21413 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
21414 (addsi3, subsi3): ... this.
21416 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
21418 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
21420 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
21422 * config/nds32/nds32.md: Adjust indention.
21424 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
21426 * config/nds32/nds32.md (feature): New attribute.
21428 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
21430 * config/nds32/nds32.md (subtype): New attribute.
21432 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
21435 * config/arm/arm-builtins.c (arm_expand_builtin): Change
21436 expansion to perform a bitwise AND of the argument followed by a
21437 boolean negation of the result.
21439 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
21441 PR rtl-optimization/84878
21442 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
21443 the basic block. Assert the use reference is not artificial and that
21444 it has an associated insn.
21446 2018-04-04 Michael Matz <matz@suse.de>
21448 * builtins.c (compute_objsize): Pass correct operand
21449 to array_at_struct_end_p.
21451 2018-04-04 Richard Biener <rguenther@suse.de>
21454 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
21455 from contexts for DINFO_LEVEL_TERSE and below.
21457 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
21459 * config/nds32/nds32-doubleword.md (move_<mode>): Require
21460 resiter_operand condition.
21461 * config/nds32/nds32.md (*move<mode>): Ditto.
21463 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
21464 Monk Chiang <sh.chiang04@gmail.com>
21466 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
21468 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
21470 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
21472 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
21473 Kito Cheng <kito.cheng@gmail.com>
21475 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
21476 nds32_cond_code_str, output_cond_branch,
21477 output_cond_branch_compare_zero, nds32_expand_cbranch,
21478 nds32_expand_cstore, nds32_expand_movcc,
21479 nds32_output_cbranchsi4_equality_zero,
21480 nds32_output_cbranchsi4_equality_reg,
21481 nds32_output_cbranchsi4_equality_reg_or_const_int,
21482 nds32_output_cbranchsi4_greater_less_zero: New functions.
21483 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
21484 nds32_expand_cstore, nds32_expand_movcc,
21485 nds32_output_cbranchsi4_equality_zero,
21486 nds32_output_cbranchsi4_equality_reg,
21487 nds32_output_cbranchsi4_equality_reg_or_const_int,
21488 nds32_output_cbranchsi4_greater_less_zero): Declare.
21489 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
21490 nds32_rimm11s_operand): New predicates.
21491 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
21492 * config/nds32/nds32.md: Rewrite all the branch and conditional move
21495 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
21497 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
21498 * config/nds32/nds32.md: Ditto.
21499 * config/nds32/pipelines.md: Ditto.
21501 2018-04-04 Richard Biener <rguenther@suse.de>
21503 PR tree-optimization/85168
21504 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
21505 propagating abnormals.
21507 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
21509 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
21511 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
21512 Kito Cheng <kito.cheng@gmail.com>
21514 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
21515 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
21516 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
21517 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
21518 * config/nds32/nds32.md (sibcall_internal): New.
21519 (sibcall_register): Remove.
21520 (sibcall_immediate): Remove.
21521 (sibcall_value_internal): New.
21522 (sibcall_value_register): Remove.
21523 (sibcall_value_immediate): Remove.
21524 * config/nds32/predicates.md (nds32_general_register_operand): New.
21525 (nds32_call_address_operand): New.
21527 2018-04-03 Jakub Jelinek <jakub@redhat.com>
21529 PR rtl-optimization/85167
21530 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
21531 bb_defs if *split_p, instead preinitialize it to NULL.
21533 PR tree-optimization/85156
21534 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
21535 evaluating the argument multiple times.
21537 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
21539 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
21541 (_mm_cvtpd_ps): Likewise.
21542 (_mm_cvttpd_epi32): Likewise.
21543 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
21544 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
21545 vector, pixel, and bool following altivec.h include.
21547 2018-04-03 Martin Sebor <msebor@redhat.com>
21549 * doc/extend.texi (Common Function Attributes): Clarify.
21550 (const attribute): Likewise.
21551 (pure attribute): Likewise.
21553 2018-04-03 Jakub Jelinek <jakub@redhat.com>
21556 * config/i386/i386.c (ix86_expand_vector_set): Use
21557 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
21559 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
21561 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
21562 instructions when changing rounding bits to preserve precision bits
21563 in the x87 control word.
21565 2018-04-03 Martin Liska <mliska@suse.cz>
21567 PR tree-optimization/82491
21568 * rtl.h (strip_offset_and_add): Replace += suboffset with
21569 poly_uint64 () + suboffset.
21571 2018-03-29 Martin Liska <mliska@suse.cz>
21572 Martin Jambor <mjambor@suse.cz>
21575 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
21576 param_type is not an integral or pointer type.
21578 2018-04-03 Richard Biener <rguenther@suse.de>
21580 * sese.h (recompute_all_dominators): Remove.
21582 2018-04-02 Martin Sebor <msebor@redhat.com>
21584 * doc/invoke.texi (-Wrestrict): Fix typos.
21586 2018-04-02 Jim Wilson <jimw@sifive.com>
21588 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
21589 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
21590 (<optab>di3, <optab>si3_extend): Likewise.
21591 (<optab>si3_mask, <optab>si3_mask_1): New.
21592 (<optab>di3_mask, <optab>di3_mask_1): New.
21593 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
21594 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
21595 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
21597 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
21599 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
21602 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
21604 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
21605 (nds32_canonicalize_comparison): New function.
21607 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
21608 Kito Cheng <kito.cheng@gmail.com>
21609 Kuan-Lin Chen <kuanlinchentw@gmail.com>
21611 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
21612 * config/nds32/constants.md (unspec_volatile_element): Add
21613 UNSPEC_VOLATILE_RELAX_GROUP.
21614 * config/nds32/nds32-relax-opt.c: New file.
21615 * config/nds32/nds32-predicates.c
21616 (nds32_symbol_load_store_p): New function.
21617 * config/nds32/nds32-protos.h
21618 (nds32_symbol_load_store_p): Declare function.
21619 (make_pass_nds32_relax_opt): Declare new rtl pass function.
21620 * config/nds32/nds32.c
21621 (nds32_register_pass): New function to register pass.
21622 (nds32_register_passes): New function to register passes.
21623 * config/nds32/nds32.md (relax_group): New pattern.
21624 * config/nds32/nds32.opt (mrelax-hint): New option.
21625 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
21627 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
21629 * config/nds32/t-nds32: Modify files dependency.
21631 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
21633 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
21634 (PROFILE_HOOK): Define its implementation.
21636 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
21638 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
21639 type and 32-bit size.
21641 2018-04-01 Jakub Jelinek <jakub@redhat.com>
21643 PR middle-end/85090
21644 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
21645 (V_128_256): New mode iterator.
21646 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
21647 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
21648 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
21650 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
21651 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
21653 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
21656 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
21657 NaN inputs correctly.
21659 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
21662 * config/rs6000/vsx.md (??r): New mode attribute.
21663 (*vsx_mov<mode>_64bit): Use it.
21664 (*vsx_mov<mode>_32bit): Likewise.
21666 2018-03-30 Martin Sebor <msebor@redhat.com>
21668 PR tree-optimization/84818
21669 * builtins.c (check_access): Use warning_n.
21671 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21674 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
21676 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
21679 2018-03-30 Julia Koval <julia.koval@intel.com>
21682 * x86-tune.def (movx, partial_reg_dependency): Enable for
21685 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
21687 PR inline-asm/84985
21688 * lra-constraints.c (process_alt_operands): Move setting
21689 this_alternative_matches below.
21691 2018-03-29 Martin Liska <mliska@suse.cz>
21694 * doc/invoke.texi: Document how LTO works with debug info.
21695 Describe auto-load support of binutils. Mention 'x86-64'
21696 as valid option value of -march option.
21698 2018-03-29 Jakub Jelinek <jakub@redhat.com>
21700 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
21703 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
21704 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
21705 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
21708 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
21711 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
21712 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
21713 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
21714 for RS6000_BTM_POWERPC64.
21715 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
21716 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
21717 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
21720 (DIVDEU): Likewise.
21722 2018-03-28 Carl Love <cel@us.ibm.com>
21725 2017-09-27 Carl Love <cel@us.ibm.com>
21727 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
21728 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
21729 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
21732 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21734 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
21735 instead of __vector bool.
21736 (_mm_max_pu8): Likewise.
21737 (_mm_min_pi16): Likewise.
21739 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
21742 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
21743 (DIVWEUO): Likewise.
21744 (DIVDEO): Likewise.
21745 (DIVDEUO): Likewise.
21746 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
21747 DIVWEUO and DIVDEUO.
21748 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
21749 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
21750 (div_extend): Likewise.
21751 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
21753 (__builtin_divweuo): Likewise.
21754 (__builtin_divdeo): Likewise.
21755 (__builtin_divdeuo): Likewise.
21757 2018-03-28 Jakub Jelinek <jakub@redhat.com>
21760 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
21761 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
21763 PR tree-optimization/82004
21764 * gimple-match-head.c (optimize_pow_to_exp): New function.
21765 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
21766 Don't fold to exp if optimize_pow_to_exp is false.
21768 2018-03-28 Martin Liska <mliska@suse.cz>
21771 * calls.c (initialize_argument_information): Fix trailing space.
21772 * common.opt: Fix typo and provide better explanation for
21773 -fsanitize-coverage option.
21774 * config/i386/i386.opt: Fix typo.
21776 2018-03-28 Jakub Jelinek <jakub@redhat.com>
21777 Martin Liska <mliska@suse.cz>
21780 * gimplify.c (asan_poison_variable): Don't do the check for
21781 gimplify_omp_ctxp here.
21782 (gimplify_decl_expr): Do it here.
21783 (gimplify_target_expr): Likewise.
21785 2018-03-28 Martin Liska <mliska@suse.cz>
21788 * config/i386/i386.c (ix86_function_arg_advance): Do not call
21789 chkp_type_bounds_count if MPX is not enabled.
21791 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
21793 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
21795 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
21798 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
21799 function to create the function decl for complex long double
21800 multiply and divide for -mabi=ieeelongdouble.
21801 (init_float128_ieee): Call it.
21803 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
21806 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
21807 -fcf-protection=branch -mibt.
21808 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
21810 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
21813 * config/arm/arm.c (arm_valid_symbolic_address): Handle
21814 arm_word_relocations.
21816 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
21819 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
21820 extern array declarations.
21822 2018-03-27 Richard Biener <rguenther@suse.de>
21824 PR middle-end/84067
21825 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
21826 explicit single_use checks.
21828 2018-03-27 Richard Biener <rguenther@suse.de>
21830 PR tree-optimization/85082
21831 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
21834 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
21836 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
21837 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
21838 Turn on fasynchronous-unwind-tables and funwind-tables.
21840 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
21843 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
21844 (*bmi_blsr_<mode>_ccz): Ditto.
21846 2018-03-26 Tom de Vries <tom@codesourcery.com>
21848 PR tree-optimization/85063
21849 * omp-general.c (offloading_function_p): New function. Factor out
21851 * omp-offload.c (pass_omp_target_link::gate): ... here.
21852 * omp-general.h (offloading_function_p): Declare.
21853 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
21854 with attribute omp declare target for offloading functions.
21856 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
21858 PR tree-optimization/84005
21859 * tree-data-ref.h (get_base_for_alignment): Declare.
21860 * tree-data-ref.c (get_base_for_alignment_1): New function.
21861 (get_base_for_alignment): Likewise.
21862 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
21863 get_base_for_alignment to find a suitable base object, instead
21864 of always using drb->base_address.
21866 2018-03-23 Jakub Jelinek <jakub@redhat.com>
21868 PR inline-asm/85022
21869 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
21870 known size by default.
21872 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
21874 PR inline-asm/85030
21875 * lra-constraints.c (process_alt_operands): Don't match BLKmode
21876 and non BLKmode operands.
21878 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21881 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
21882 Clean up attributes.
21884 2018-03-23 Richard Biener <rguenther@suse.de>
21887 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
21888 we are going to emit early debug for LTO.
21890 2018-03-23 Jakub Jelinek <jakub@redhat.com>
21892 PR inline-asm/85034
21893 * function.c (match_asm_constraints_1): Don't optimize if input
21894 doesn't satisfy general_operand predicate for output's mode.
21896 PR inline-asm/85022
21897 * alias.c (write_dependence_p): Don't require for x_canonicalized
21898 non-VOIDmode if x has VOIDmode.
21901 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
21902 just don't try to optimize it rather than assert it never happens.
21904 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
21906 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
21907 macro expansions for definition of ST_INTERNAL_<mode> and
21908 LD_INTERNAL_<mode> builtins.
21909 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
21911 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
21913 (altivec_expand_st_builtin): Likewise.
21914 (altivec_expand_builtin): Remove calls to deleted functions.
21915 (rs6000_address_for_altivec): Delete this function.
21916 * config/rs6000/vector.md: Remove expands for
21917 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
21919 2018-03-22 Sudakshina Das <sudi.das@arm.com>
21922 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
21923 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
21924 re-computing once computed.
21925 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
21926 (arm_init_machine_status): Initialize
21927 machine->static_chain_stack_bytes.
21929 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
21932 * doc/extend.texi: Add four new prototypes for vec_ld.
21933 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
21934 definitions for more logical presentation.
21935 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
21936 entries for V1TI variants of __builtin_altivec_ld builtin.
21937 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
21938 handling of V1TI variant of LVX icode pattern.
21939 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
21940 (rs6000_gimple_fold_builtin): Likewise.
21941 (altivec_init_builtins): Add code to define
21942 __builtin_altivec_lvx_v1ti function.
21944 2018-03-22 Jakub Jelinek <jakub@redhat.com>
21946 PR inline-asm/84941
21947 * function.c (match_asm_constraints_1): Don't do the optimization
21948 if input isn't a REG, SUBREG, MEM or constant.
21950 2018-03-22 Tom de Vries <tom@codesourcery.com>
21952 PR tree-optimization/84956
21953 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
21954 bb_has_abnormal_pred.
21956 2018-03-22 Jakub Jelinek <jakub@redhat.com>
21959 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
21960 DECL_INITIAL (decl) to decl at the end.
21961 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
21962 adjust the comment.
21964 2018-03-21 Joseph Myers <joseph@codesourcery.com>
21966 * doc/extend.texi (__builtin_tgmath): Document when complex
21967 integer types are treated as _Complex _Float64.
21969 2018-03-21 Tom de Vries <tom@codesourcery.com>
21971 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
21973 2018-03-21 Jakub Jelinek <jakub@redhat.com>
21975 PR tree-optimization/84960
21976 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
21977 if it is ENTRY block, move them into single succ of ENTRY in that case.
21979 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
21981 PR tree-optimization/84811
21982 * poly-int.h (poly_span_traits): Remove the T3 parameter and
21983 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
21984 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
21985 (known_subrange_p): Update accordingly. Cast each value involved
21986 in the size comparison, rather than casting the result of the
21989 2018-03-21 Jakub Jelinek <jakub@redhat.com>
21991 PR tree-optimization/84982
21992 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
21993 by flipping the least significant bit rather than all bits from
21994 bitpos to bitpos + bitsize - 1.
21996 2018-03-21 Nathan Sidwell <nathan@acm.org>
21998 * doc/extend.texi (Deprecated Features): Remove mention of
21999 long-deleted deprecations.
22001 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22004 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
22005 * configure: Regenerate.
22007 2018-03-21 Tom de Vries <tom@codesourcery.com>
22009 PR tree-optimization/83126
22010 * tree-parloops.c (num_phis): New function.
22011 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
22013 2018-03-21 Nathan Sidwell <nathan@acm.org>
22015 * doc/extend.texi (Deprecated Features): Update deprecated flags,
22016 mention anon-struct/union members and trailing attributes.
22018 2018-03-21 Bin Cheng <bin.cheng@arm.com>
22020 PR tree-optimization/84969
22021 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
22022 builtin memset partitions if they set different rhs values.
22024 2018-03-21 Jakub Jelinek <jakub@redhat.com>
22026 PR rtl-optimization/84989
22027 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
22028 VEC_DUPLICATE with scalar result mode.
22030 2018-03-21 Martin Liska <mliska@suse.cz>
22033 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
22034 not intended return statement.
22036 2018-03-21 Martin Liska <mliska@suse.cz>
22039 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
22040 (chkp_find_bound_slots_1): Limit number of iterations.
22042 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
22045 * Minor grammar fixes for x86 options.
22047 2018-03-20 Jakub Jelinek <jakub@redhat.com>
22050 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
22051 holding REG_CFA_RESTORE notes, instead turn them into a USE.
22053 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
22056 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
22057 (altivec_lvx_<mode>_1op): Likewise.
22058 (altivec_stvx_<mode>_2op): Likewise.
22059 (altivec_stvx_<mode>_1op): Likewise.
22060 (altivec_lvx_<VM2:mode>): New define_expand.
22061 (altivec_stvx_<VM2:mode>): Likewise.
22062 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
22063 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
22064 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
22065 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
22066 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
22067 (rs6000_gen_lvx): Likewise.
22068 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
22069 (altivec_expand_stv_builtin): Likewise.
22070 (altivec_expand_builtin): Likewise.
22071 * config/rs6000/vector.md: Likewise.
22073 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22076 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
22079 2018-03-20 Richard Biener <rguenther@suse.de>
22082 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
22083 sign-conversions as zero, fall back to standard scalar_stmt
22086 2018-03-20 Martin Liska <mliska@suse.cz>
22089 * predict.c (rebuild_frequencies): Handle case when we have
22090 PROFILE_ABSENT, but flag_guess_branch_prob is false.
22092 2018-03-20 Jakub Jelinek <jakub@redhat.com>
22095 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
22096 flag_section_anchors.
22097 * varasm.c (use_blocks_for_decl_p): Remove hack for
22098 dw2_force_const_mem.
22101 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
22103 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
22104 be created, use lowpart_subreg of operands[0] rather than operands[0]
22106 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
22107 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
22108 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
22109 and n constraint instead of aarch64_shift_imm_di and Usd.
22110 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
22111 (*aarch64_<optab>_reg_minus<mode>3): ... this.
22113 2018-03-20 Sudakshina Das <sudi.das@arm.com>
22116 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
22117 to favor GPR over NEON registers.
22118 (<shift>di3_neon): Likewise.
22120 2018-03-20 Tom de Vries <tom@codesourcery.com>
22123 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
22124 (nvptx_process_pars): Emit bar.sync asap and alap.
22126 2018-03-20 Tom de Vries <tom@codesourcery.com>
22129 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
22130 seen_label if seen_label is already set.
22132 2018-03-20 Jakub Jelinek <jakub@redhat.com>
22135 * config/i386/i386.c (fold_builtin_cpu): For features above 31
22136 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
22137 Use 1U instead of 1. Formatting fixes.
22140 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
22141 instead of TREE_TYPE (s1) for the return value.
22143 2018-03-19 Jakub Jelinek <jakub@redhat.com>
22145 PR tree-optimization/84946
22146 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
22147 bitsize + bitsize in poly_uint64 rather than poly_int64.
22150 * dwarf2asm.c: Include fold-const.c.
22151 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
22152 of decl rather than decl itself.
22154 PR rtl-optimization/84643
22155 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
22157 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
22160 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
22161 calling assemble_variable.
22163 2018-03-19 Sudakshina Das <sudi.das@arm.com>
22166 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
22167 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
22169 2018-03-19 Jim Wilson <jimw@sifive.com>
22172 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
22173 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
22174 (riscv_first_stack_step): Likewise.
22175 (riscv_option_override): Use STACK_BOUNDARY instead of
22176 MIN_STACK_BOUNDARY.
22177 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
22178 MIN_STACK_BOUNDARY.
22179 (BIGGEST_ALIGNMENT): Set to 128.
22180 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
22181 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
22184 2018-03-19 Richard Biener <rguenther@suse.de>
22186 PR tree-optimization/84933
22187 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
22188 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
22190 2018-03-19 Richard Biener <rguenther@suse.de>
22192 PR tree-optimization/84859
22193 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
22194 (cond_if_else_store_replacement): Perform sinking operation on
22195 single-store BBs regardless of MAX_STORES_TO_SINK setting.
22196 Generalize what a BB with a single eligible store is.
22198 2018-03-19 Richard Biener <rguenther@suse.de>
22200 PR tree-optimization/84929
22201 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
22202 chrec_is_positive against non-chrec arg.
22204 2018-03-19 Tamar Christina <tamar.christina@arm.com>
22207 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
22209 2018-03-18 Martin Liska <mliska@suse.cz>
22211 PR rtl-optimization/84635
22212 * regrename.c (build_def_use): Use matches_mode only when
22215 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
22217 PR tree-optimization/84913
22218 * tree-vect-loop.c (vectorizable_reduction): Don't try to
22219 vectorize chains of COND_EXPRs.
22221 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
22223 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
22225 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
22227 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
22229 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
22231 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
22233 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
22234 Kito Cheng <kito.cheng@gmail.com>
22236 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
22237 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
22238 (nds32_adjust_reg_alloc_order): New function.
22239 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
22241 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
22243 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
22244 nds32_print_operand, nds32_print_operand_address): Use
22245 HOST_WIDE_INT_PRINT_DEC instead.
22247 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
22249 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
22251 2018-03-17 Jakub Jelinek <jakub@redhat.com>
22254 * config/i386/i386.c (initial_ix86_tune_features,
22255 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
22256 unsigned long long.
22257 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
22258 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
22259 rather than 1u << ix86_tune. Formatting fix.
22260 (ix86_option_override_internal): Change ix86_arch_mask from
22261 unsigned int to unsigned HOST_WIDE_INT, initialize to
22262 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
22263 (ix86_function_specific_restore): Likewise.
22265 2018-03-16 Jakub Jelinek <jakub@redhat.com>
22268 * postreload.c (reload_combine_recognize_pattern): Perform
22269 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
22270 truncate_int_for_mode the result for the destination's mode.
22273 * hsa-gen.c (mem_type_for_type): Fix comment typo.
22274 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
22276 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
22279 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
22282 * lra-assigns.c (lra_split_hard_reg_for): Don't use
22283 regno_allocno_class_array and sorted_pseudos.
22284 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
22285 insns where regno is used.
22287 2018-03-16 Martin Liska <mliska@suse.cz>
22290 * multiple_target.c (create_dispatcher_calls): Redirect
22291 reference in the symbol table.
22293 2018-03-16 Martin Liska <mliska@suse.cz>
22296 * multiple_target.c (create_dispatcher_calls): Redirect also
22299 2018-03-16 Jakub Jelinek <jakub@redhat.com>
22303 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
22304 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
22305 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
22307 2018-03-16 Julia Koval <julia.koval@intel.com>
22309 * doc/invoke.texi (Skylake Server): Add CLWB.
22310 Cannonlake): Remove CLWB.
22312 2018-03-16 Jakub Jelinek <jakub@redhat.com>
22314 PR tree-optimization/84841
22315 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
22317 (FLOAT_ONE_CONST_TYPE): Define.
22318 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
22319 (sort_by_operand_rank): Put entries with higher constant_type last
22320 rather than first to match comments.
22322 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
22324 * config/nios2/nios2.md (movsi_internal): Fix thinko in
22327 2018-03-15 Jakub Jelinek <jakub@redhat.com>
22330 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
22331 check and use address of target always.
22333 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
22336 * config/i386/i386.c (indirect_thunk_needed): Update comments.
22337 (indirect_thunk_bnd_needed): Likewise.
22338 (indirect_thunks_used): Likewise.
22339 (indirect_thunks_bnd_used): Likewise.
22340 (indirect_return_needed): New.
22341 (indirect_return_bnd_needed): Likewise.
22342 (output_indirect_thunk_function): Add a bool argument for
22344 (output_indirect_thunk_function): Don't generate alias for
22345 function return thunk.
22346 (ix86_code_end): Call output_indirect_thunk_function to generate
22347 function return thunks.
22348 (ix86_output_function_return): Set indirect_return_bnd_needed
22349 and indirect_return_needed instead of indirect_thunk_bnd_needed
22350 and indirect_thunk_needed.
22352 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
22354 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
22355 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
22356 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
22358 2018-03-15 David Malcolm <dmalcolm@redhat.com>
22359 Paul Hua <paul.hua.gm@gmail.com>
22362 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
22364 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
22366 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
22367 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
22368 resp. SFmode cases.
22370 2018-03-15 Tamar Christina <tamar.christina@arm.com>
22373 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
22374 instead of GET_MODE_SIZE when comparing Units.
22376 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
22379 * varasm.c (hash_section): Return an unchangeble hash value
22380 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
22381 Return !aarch64_can_use_per_function_literal_pools_p ().
22383 2018-03-15 Jakub Jelinek <jakub@redhat.com>
22386 * optabs.c (emit_conditional_move): Pass address of cmode's copy
22387 rather than address of cmode as last argument to prepare_cmp_insn.
22389 2018-03-15 Julia Koval <julia.koval@intel.com>
22391 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
22392 F_AVX512VNNI, F_AVX512BITALG): New.
22394 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
22397 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
22398 insn for floating-point loads and stores.
22400 2018-03-14 Carl Love <cel@us.ibm.com>
22402 * config/rs6000/rs6000-c.c: Add macro definitions for
22403 ALTIVEC_BUILTIN_VEC_PERMXOR.
22404 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
22405 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
22406 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
22407 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
22409 * config/doc/extend.texi: Add prototypes for vec_permxor.
22411 2018-03-14 David Malcolm <dmalcolm@redhat.com>
22414 * diagnostic-show-locus.c (class layout_point): Convert m_line
22415 from int to linenum_type.
22416 (line_span::comparator): Use linenum "compare" function when
22417 comparing line numbers.
22418 (test_line_span): New function.
22419 (layout_range::contains_point): Convert param "row" from int to
22421 (layout_range::intersects_line_p): Likewise.
22422 (layout::will_show_line_p): Likewise.
22423 (layout::print_source_line): Likewise.
22424 (layout::should_print_annotation_line_p): Likewise.
22425 (layout::print_annotation_line): Likewise.
22426 (layout::print_leading_fixits): Likewise.
22427 (layout::annotation_line_showed_range_p): Likewise.
22428 (struct line_corrections): Likewise for field m_row.
22429 (line_corrections::line_corrections): Likewise for param "row".
22430 (layout::print_trailing_fixits): Likewise.
22431 (layout::get_state_at_point): Likewise.
22432 (layout::get_x_bound_for_row): Likewise.
22433 (layout::print_line): Likewise.
22434 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
22435 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
22436 * input.c (selftest::test_linenum_comparisons): New function.
22437 (selftest::input_c_tests): Call it.
22438 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
22439 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
22440 * selftest.h (ASSERT_GT): New macro.
22441 (ASSERT_GT_AT): New macro.
22442 (ASSERT_LT): New macro.
22443 (ASSERT_LT_AT): New macro.
22445 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
22447 PR rtl-optimization/84780
22448 * combine.c (distribute_links): Don't make a link based on pc_rtx.
22450 2018-03-14 Martin Liska <mliska@suse.cz>
22452 * tree.c (record_node_allocation_statistics): Use
22453 get_stats_node_kind.
22454 (get_stats_node_kind): New function extracted from
22455 record_node_allocation_statistics.
22456 (free_node): Use get_stats_node_kind.
22458 2018-03-14 Richard Biener <rguenther@suse.de>
22460 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
22461 that the value-set of ANTIC_IN doesn't grow.
22464 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
22466 (BB_VISITED_WITH_VISITED_SUCCS): New define.
22467 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
22469 2018-03-14 Julia Koval <julia.koval@intel.com>
22471 * config.gcc (icelake-client, icelake-server): New.
22473 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
22474 (initial_ix86_arch_features): Ditto.
22475 (PTA_SKYLAKE): Add SGX.
22476 (PTA_ICELAKE): Remove.
22477 (PTA_ICELAKE_CLIENT): New.
22478 (PTA_ICELAKE_SERVER): New.
22479 (ix86_option_override_internal): Split up icelake on icelake client and
22481 (get_builtin_code_for_version): Ditto.
22482 (fold_builtin_cpu): Ditto.
22483 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
22484 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
22485 * config/i386/i386.h (processor_type): Ditto.
22486 * doc/invoke.texi: Ditto.
22488 2018-03-14 Jakub Jelinek <jakub@redhat.com>
22491 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
22492 INTEGER_CST offset, add it together with bitpos / 8 and
22493 sign extend based on POINTER_SIZE.
22497 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
22500 * config/i386/constraints.md (Yc): New register constraint.
22501 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
22502 Use Yc constraint for alternative 2 of operand 0. Remove
22503 preferred_for_speed attribute.
22505 2018-03-14 Richard Biener <rguenther@suse.de>
22507 PR tree-optimization/84830
22508 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
22509 with the old one to avoid oscillations.
22511 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
22514 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
22516 (assign_by_spills): Return a flag of reload assignment failure.
22517 Do not process the reload assignment failures. Do not spill other
22518 reload pseudos if they has the same reg class. Update n if
22520 (lra_assign): Add a return arg. Set up from the result of
22521 assign_by_spills call.
22522 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
22523 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
22524 usage_insns if it is not NULL.
22525 (spill_hard_reg_in_range): New function.
22526 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
22527 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
22528 function prototypes.
22529 (lra_assign): Change prototype.
22530 * lra.c (lra): Add code to deal with fails by splitting hard reg
22533 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
22535 * config/riscv/riscv.opt (mrelax): New option.
22536 * config/riscv/riscv.c (riscv_file_start): Emit ".option
22537 "norelax" when riscv_mrelax is disabled.
22538 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
22540 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22543 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
22544 reassociation for int modes.
22546 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
22548 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
22549 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
22551 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
22552 * config/aarch64/aarch64-sve.md
22553 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
22554 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
22555 (*extend<mode><Vwide>2): Rename to...
22556 (aarch64_sve_extend<mode><Vwide>2): ...this.
22557 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
22558 renaming the old pattern to...
22559 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
22561 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
22562 define_expand, renaming the old pattern to...
22563 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
22564 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
22565 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
22566 account when deciding which SVE instruction the optab should use.
22567 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
22569 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
22571 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
22572 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
22573 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
22574 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
22575 (tlsdesc_small_<mode>): Turn a define_expand and use
22576 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
22577 (tlsdesc_small_advsimd_<mode>): ...this.
22578 (tlsdesc_small_sve_<mode>): New pattern.
22580 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
22582 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
22583 (UNSPEC_UMUL_HIGHPART): New constants.
22584 (MUL_HIGHPART): New int iteraor.
22585 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
22586 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
22588 (*<su>mul<mode>3_highpart): New define_insn.
22590 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
22593 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
22596 2018-03-13 Martin Liska <mliska@suse.cz>
22599 * (sem_item_optimizer::sem_item_optimizer): Initialize new
22601 (sem_item_optimizer::~sem_item_optimizer): Release it.
22602 (sem_item_optimizer::merge_classes): Register variable aliases.
22603 (sem_item_optimizer::fixup_pt_set): New function.
22604 (sem_item_optimizer::fixup_points_to_sets): Likewise.
22605 * ipa-icf.h: Declare new variables and functions.
22607 2018-03-13 Jakub Jelinek <jakub@redhat.com>
22609 PR middle-end/84834
22610 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
22611 integer_pow2p@2 and test integer_pow2p in condition.
22612 (A < 0 ? C : 0): Similarly for @1.
22614 PR middle-end/84831
22615 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
22616 characters starting at p contain '\0' character, don't look beyond
22620 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
22621 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
22624 * reg-stack.c (change_stack): Change update_end var from int to
22625 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
22626 also call set_block_for_insn on the newly added insns and rescan.
22629 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
22630 on the last operand.
22633 * tree.c (stabilize_reference_1): Return save_expr (e) for
22634 STATEMENT_LIST even if it doesn't have side-effects.
22636 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
22638 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
22640 2018-03-12 Renlin Li <renlin.li@arm.com>
22642 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
22643 aarch64_output_scalar_simd_mov_immediate.
22645 2018-03-12 Martin Sebor <msebor@redhat.com>
22647 PR tree-optimization/83456
22648 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
22649 for perfectly overlapping calls to memcpy.
22650 (gimple_fold_builtin_memory_chk): Same.
22651 (gimple_fold_builtin_strcpy): Handle no-warning.
22652 (gimple_fold_builtin_stxcpy_chk): Same.
22653 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
22655 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
22657 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
22658 parameter. Use it for SFmode.
22659 (rs6000_function_arg_advance_1): Adjust.
22660 (rs6000_function_arg): Adjust.
22661 (rs6000_gimplify_va_arg): Pass false for that new parameter.
22663 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
22665 PR rtl-optimization/84169
22666 PR rtl-optimization/84780
22667 * combine.c (can_combine_p): Check for a 2-insn combination whether
22668 the destination register is used between the two insns, too.
22670 2018-03-12 Richard Biener <rguenther@suse.de>
22672 PR tree-optimization/84803
22673 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
22674 for refs DR analysis didn't process.
22676 2018-03-12 Richard Biener <rguenther@suse.de>
22678 PR tree-optimization/84777
22679 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
22680 force-vectorize loops ignore whether we are optimizing for size.
22682 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
22684 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
22685 (TARGET_MD_ASM_ADJUST): Define.
22687 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
22688 Kito Cheng <kito.cheng@gmail.com>
22689 Chung-Ju Wu <jasonwucj@gmail.com>
22691 * config/nds32/nds32.c (nds32_compute_stack_frame,
22692 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
22693 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
22694 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
22695 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
22696 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
22697 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
22698 * config/nds32/nds32.md (prologue, epilogue): Use macro
22699 NDS32_V3PUSH_AVAILABLE_P to do checking.
22701 2018-03-11 Jakub Jelinek <jakub@redhat.com>
22704 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
22705 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
22706 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
22707 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
22708 addition of most attributes on !orig_type_die or the attribute not
22709 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
22711 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
22712 Chung-Ju Wu <jasonwucj@gmail.com>
22714 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
22715 __NDS32_VH__ macro.
22716 * config/nds32/nds32.opt (mvh): New option.
22718 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
22719 Chung-Ju Wu <jasonwucj@gmail.com>
22721 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
22723 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
22724 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
22727 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
22728 Chung-Ju Wu <jasonwucj@gmail.com>
22730 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
22732 * config/nds32/nds32-multiple.md (strlensi): New pattern.
22733 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
22735 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
22736 Kito Cheng <kito.cheng@gmail.com>
22737 Chung-Ju Wu <jasonwucj@gmail.com>
22739 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
22740 UNSPEC_FFMISM and UNSPEC_FLMISM.
22741 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
22742 for ffb, ffmism and flmism.
22743 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
22744 (unspec_ffmism): Ditto.
22745 (unspec_flmism): Ditto.
22746 (nds32_expand_builtin_impl): Check if string extension is available.
22747 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
22748 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
22750 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
22753 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
22756 * lra-assigns.c (assign_by_spills): Return a flag of reload
22757 assignment failure. Do not process the reload assignment
22758 failures. Do not spill other reload pseudos if they has the same
22760 (lra_assign): Add a return arg. Set up from the result of
22761 assign_by_spills call.
22762 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
22763 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
22764 usage_insns if it is not NULL.
22765 (spill_hard_reg_in_range): New function.
22766 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
22767 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
22768 function prototypes.
22769 (lra_assign): Change prototype.
22770 * lra.c (lra): Add code to deal with fails by splitting hard reg
22773 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
22776 * config/i386/i386.opt: Replace Enforcment with Enforcement.
22778 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
22781 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
22782 (dw_val_node): Add val_symbolic_view.
22783 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
22784 (symview_upper_bound): New.
22785 (new_line_info_table): Initialize symviews_since_reset.
22786 (dwarf2out_source_line): Count symviews_since_reset and set
22787 symview_upper_bound.
22788 (dw_val_equal_p): Handle symview.
22789 (add_AT_symview): New.
22790 (print_dw_val): Handle symview.
22791 (attr_checksum, attr_checksum_ordered): Likewise.
22792 (same_dw_val_p, size_of_die): Likewise.
22793 (value_format, output_die): Likewise.
22794 (add_high_low_attributes): Use add_AT_symview for entry_view.
22795 (dwarf2out_finish): Reset symview_upper_bound, clear
22798 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
22801 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
22802 Add strict argument and use it.
22803 (rs6000_split_multireg_move): Update for new strict argument.
22804 (mem_operand_gpr): Disallow all non-offsettable addresses.
22805 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
22807 2018-03-09 Jakub Jelinek <jakub@redhat.com>
22810 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
22811 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
22812 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
22815 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
22816 decl, use remap_type if we want to use the type.
22818 2018-03-09 Martin Sebor <msebor@redhat.com>
22820 PR tree-optimization/84526
22821 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
22823 (builtin_access::generic_overlap): Be prepared to handle non-array
22826 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
22828 PR rtl-optimization/84682
22829 * lra-constraints.c (process_address_1): Check is_address flag
22830 for address constraints.
22831 (process_alt_operands): Likewise.
22832 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
22833 preprocess_constraints.
22834 * recog.h (preprocess_constraints): Add oploc parameter.
22836 * recog.c (preprocess_constraints): Test address_operand for
22837 CT_ADDRESS constraints.
22839 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
22842 * lra-assigns.c (assign_by_spills): Return a flag of reload
22843 assignment failure. Do not process the reload assignment
22844 failures. Do not spill other reload pseudos if they has the same
22846 (lra_assign): Add a return arg. Set up from the result of
22847 assign_by_spills call.
22848 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
22849 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
22850 usage_insns if it is not NULL.
22851 (spill_hard_reg_in_range): New function.
22852 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
22853 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
22854 function prototypes.
22855 (lra_assign): Change prototype.
22856 * lra.c (lra): Add code to deal with fails by splitting hard reg
22859 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22862 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
22863 Accept complain bool parameter. Only emit errors if it is true.
22864 (arm_parse_cpu_option_name): Likewise.
22865 (arm_target_thumb_only): Adjust callers of the above.
22866 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
22867 prototype to take a default true bool parameter.
22868 (arm_parse_arch_option_name): Likewise.
22870 2018-03-09 David Malcolm <dmalcolm@redhat.com>
22871 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
22875 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
22876 * configure: Regenerate.
22877 * configure.ac ("linker --version-script option"): New.
22878 ("linker soname option"): New.
22880 2018-03-09 Richard Biener <rguenther@suse.de>
22882 PR tree-optimization/84775
22883 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
22884 immediate uses of predicate stmts and mark them modified.
22887 PR tree-optimization/84178
22888 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
22890 (version_loop_for_if_conversion): Delay update_ssa call.
22891 (tree_if_conversion): Delay update_ssa until after predicate
22894 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
22897 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
22898 when the function accesses prior frames.
22900 2018-03-08 Jakub Jelinek <jakub@redhat.com>
22903 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
22904 gen_llsym, otherwise call maybe_gen_llsym.
22906 PR inline-asm/84742
22907 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
22908 has ',' character inside of it.
22910 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22913 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
22914 as clobbering CC_REGNUM.
22916 2018-03-08 Richard Biener <rguenther@suse.de>
22918 PR middle-end/84552
22919 * tree-scalar-evolution.c: Include tree-into-ssa.h.
22920 (follow_copies_to_constant): Do not follow SSA names registered
22923 2018-03-08 Richard Biener <rguenther@suse.de>
22925 PR tree-optimization/84178
22926 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
22928 (version_loop_for_if_conversion): Delay update_ssa call.
22929 (tree_if_conversion): Delay update_ssa until after predicate
22932 2018-03-08 David Malcolm <dmalcolm@redhat.com>
22934 PR tree-optimization/84178
22935 * tree-if-conv.c (release_bb_predicate): Remove the
22936 the assertion that the stmts have NULL use_ops.
22937 Discard the statements, asserting that they haven't
22938 yet been added to a BB.
22940 2018-03-08 Richard Biener <rguenther@suse.de>
22942 PR tree-optimization/84746
22943 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
22944 (phi_translate): Pass in destination ANTIC_OUT set.
22945 (phi_translate_1): Likewise. For a simplified result lookup
22946 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
22947 (phi_translate_set): Adjust.
22948 (do_pre_regular_insertion): Likewise.
22949 (do_pre_partial_partial_insertion): Likewise.
22951 2018-03-08 Martin Liska <mliska@suse.cz>
22953 PR gcov-profile/84735
22954 * doc/gcov.texi: Document usage of profile files.
22955 * gcov-io.h: Document changes in the format.
22957 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
22961 * dwarf2out.c (struct dw_line_info_table): Update comments for
22963 (FORCE_RESET_NEXT_VIEW): New.
22964 (FORCE_RESETTING_VIEW_P): New.
22965 (RESETTING_VIEW_P): Check for -1 too.
22966 (ZERO_VIEW_P): Likewise.
22967 (new_line_info_table): Force-reset next view.
22968 (dwarf2out_begin_function): Likewise.
22969 (dwarf2out_source_line): Simplify zero_view_p initialization.
22970 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
22971 view directly. Omit view when omitting .loc at line 0.
22973 2018-03-08 Jakub Jelinek <jakub@redhat.com>
22975 PR tree-optimization/84740
22976 * tree-switch-conversion.c (process_switch): Call build_constructors
22977 only if info.phi_count is non-zero.
22979 PR tree-optimization/84739
22980 * tree-tailcall.c (find_tail_calls): Check call arguments against
22981 DECL_ARGUMENTS (current_function_decl) rather than
22982 DECL_ARGUMENTS (func) when checking for tail recursion.
22984 2018-03-07 Jakub Jelinek <jakub@redhat.com>
22986 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
22987 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
22988 Volker Reichelt's entry and add entries for people that perform
22989 GCC fuzzy testing and report numerous bugs.
22991 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
22994 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
22995 readonly data in sdata, if that is disabled.
22996 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
22997 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
22998 -mreadonly-in-sdata option.
23000 2018-03-07 Martin Sebor <msebor@redhat.com>
23002 PR tree-optimization/84468
23003 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
23004 basic block when looking for nul assignment.
23006 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
23009 * except.h (output_function_exception_table): Adjust prototype.
23010 * except.c (output_function_exception_table): Remove FNNAME parameter
23011 and add SECTION parameter. Ouput one part of the table at a time.
23012 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
23013 the first part of the exception table and emit unwind directives.
23014 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
23015 (i386_pe_seh_cold_init): Likewise.
23016 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
23017 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
23018 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
23019 (ix86_output_call_insn): Emit a nop in one more case for SEH.
23020 * config/i386/winnt.c: Include except.h.
23021 (struct seh_frame_state): Add reg_offset, after_prologue and
23022 in_cold_section fields.
23023 (i386_pe_seh_end_prologue): Set seh->after_prologue.
23024 (i386_pe_seh_cold_init): New function.
23025 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
23026 to seh->in_cold_section.
23027 (seh_emit_push): Record the offset of the push.
23028 (seh_emit_save): Record the offet of the save.
23029 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
23030 Test seh->after_prologue to disregard the epilogue.
23031 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
23032 (i386_pe_end_cold_function): New function.
23034 2018-03-07 Jakub Jelinek <jakub@redhat.com>
23037 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
23038 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
23041 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
23043 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
23044 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
23046 PR middle-end/84723
23047 * multiple_target.c: Include tree-inline.h and intl.h.
23048 (expand_target_clones): Diagnose and fail if node->definition and
23049 !tree_versionable_function_p (node->decl).
23051 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
23053 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
23055 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
23056 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
23057 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
23059 2018-03-06 Jakub Jelinek <jakub@redhat.com>
23062 * combine.c (try_combine): Use reg_or_subregno instead of handling
23063 just paradoxical SUBREGs and REGs.
23065 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
23067 * config/arc/arc.c (arc_finalize_pic): Remove function.
23068 (arc_must_save_register): We use single base PIC register, remove
23069 checks to save/restore the PIC register.
23070 (arc_expand_prologue): Likewise.
23071 * config/arc/arc-protos.h (arc_set_default_type_attributes):
23073 (arc_verify_short): Likewise.
23074 (arc_attr_type): Likewise.
23075 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
23076 (walk_stores): Likewise.
23077 (arc_address_cost): Make it static.
23078 (arc_verify_short): Likewise.
23079 (branch_dest): Likewise.
23080 (arc_attr_type): Likewise.
23081 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
23082 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
23083 (arc_final_prescan_insn): Remove inserting the nops due to
23084 hardware hazards. It is done in reorg step.
23085 (insn_length_variant_t): Remove.
23086 (insn_length_parameters_t): Likewise.
23087 (arc_insn_length_parameters): Likewise.
23088 (arc_get_insn_variants): Likewise.
23089 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
23091 2018-03-06 Jakub Jelinek <jakub@redhat.com>
23093 PR inline-asm/84683
23094 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
23097 PR tree-optimization/84687
23098 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
23100 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
23102 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23104 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
23105 Rename to ppc_speculation_barrier.
23106 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
23107 __builtin_ppc_speculation_barrier.
23109 2018-03-05 Jakub Jelinek <jakub@redhat.com>
23112 * combine.c (combine_simplify_rtx): Don't try to simplify if
23113 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
23116 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
23118 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
23119 to 32 bytes when compiling for POWER9.
23121 2018-03-05 Jakub Jelinek <jakub@redhat.com>
23124 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
23125 regparm >= 3 with no arg reg available also for calls with
23126 flag_force_indirect_call. Pass decl to ix86_function_regparm.
23129 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
23131 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
23133 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
23136 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
23138 2018-03-05 Richard Biener <rguenther@suse.de>
23140 PR tree-optimization/84486
23141 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
23142 When inserting a __builtin_assume_aligned call set the LHS
23143 SSA name alignment info accordingly.
23145 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
23147 PR tree-optimization/84114
23148 * config/aarch64/aarch64.c (aarch64_reassociation_width)
23149 Avoid reassociation of FLOAT_MODE addition.
23151 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
23153 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
23154 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
23155 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
23156 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
23157 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
23158 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
23159 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
23161 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
23162 __builtin_ia32_wbinvd): New builtins.
23163 (SPECIAL_ARGS2): New.
23164 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
23165 (SPECIAL_ARGS2): New.
23166 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
23167 (ix86_valid_target_attribute_inner_p): Ditto.
23168 (ix86_init_mmx_sse_builtins): Add special_args2.
23169 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
23170 TARGET_WBNOINVD_P): New.
23171 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
23172 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
23173 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
23174 * config/i386/immintrin.h (_wbinvd): New intrinsic.
23175 * config/i386/pconfigintrin.h: New file.
23176 * config/i386/wbnoinvdintrin.h: Ditto.
23177 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
23179 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
23181 2018-03-05 Richard Biener <rguenther@suse.de>
23183 PR tree-optimization/84670
23184 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
23186 (BB_VISITED_WITH_VISITED_SUCCS): New define.
23187 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
23188 (compute_antic_aux): Only assert the number of values in ANTIC_IN
23189 doesn't grow if all successors (recursively) were visited at least
23192 2018-03-05 Richard Biener <rguenther@suse.de>
23194 PR tree-optimization/84650
23195 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
23196 if executed in the loop pipeline.
23198 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
23200 * doc/configfiles.texi (Configuration Files): Move info about
23201 conditionalizing $target-protos.h to...
23202 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
23203 differs from $target-protos.h.
23205 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
23206 Chung-Ju Wu <jasonwucj@gmail.com>
23208 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
23209 * config/nds32/nds32-multiple.md (setmemsi): Define.
23210 * config/nds32/nds32-memory-manipulation.c
23211 (nds32_gen_dup_4_byte_to_word_value): New.
23212 (emit_setmem_word_loop): New.
23213 (emit_setmem_byte_loop): New.
23214 (nds32_expand_setmem_loop): New.
23215 (nds32_expand_setmem_loop_v3m): New.
23216 (nds32_expand_setmem_unroll): New.
23217 (nds32_expand_setmem): New.
23219 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
23220 Chung-Ju Wu <jasonwucj@gmail.com>
23222 * config/nds32/nds32-memory-manipulation.c
23223 (nds32_emit_load_store): New.
23224 (nds32_emit_post_inc_load_store): New.
23225 (nds32_emit_mem_move): New.
23226 (nds32_emit_mem_move_block): New.
23227 (nds32_expand_movmemsi_loop_unknown_size): New.
23228 (nds32_expand_movmemsi_loop_known_size): New.
23229 (nds32_expand_movmemsi_loop): New.
23230 (nds32_expand_movmemsi_unroll): New.
23231 (nds32_expand_movmemqi): Rename ...
23232 (nds32_expand_movmemsi): ... to this.
23233 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
23234 (movmemsi): ... to this.
23235 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
23236 (nds32_expand_movmemsi): ... to this.
23238 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
23239 Monk Chiang <sh.chiang04@gmail.com>
23240 Chung-Ju Wu <jasonwucj@gmail.com>
23242 * config/nds32/nds32-protos.h
23243 (nds32_expand_load_multiple): New arguments.
23244 (nds32_expand_store_multiple): Ditto.
23245 (nds32_valid_multiple_load_store): Rename ...
23246 (nds32_valid_multiple_load_store_p): ... to this.
23247 * config/nds32/nds32-memory-manipulation.c
23248 (nds32_expand_load_multiple): Refine implementation.
23249 (nds32_expand_store_multiple): Ditto.
23250 * config/nds32/nds32-multiple.md
23251 (load_multiple): Update nds32_expand_load_multiple interface.
23252 (store_multiple): Update nds32_expand_store_multiple interface.
23253 * config/nds32/nds32-predicates.c
23254 (nds32_valid_multiple_load_store): Rename ...
23255 (nds32_valid_multiple_load_store_p): ... to this and refine
23257 * config/nds32/predicates.md
23258 (nds32_load_multiple_and_update_address_operation): New predicate.
23259 (nds32_store_multiple_and_update_address_operation): New predicate.
23261 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
23262 Chung-Ju Wu <jasonwucj@gmail.com>
23264 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
23265 (combo): New attribute.
23266 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
23268 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
23270 * config/nds32/nds32.opt: Change -mcmodel= default value.
23272 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
23273 Monk Chiang <sh.chiang04@gmail.com>
23274 Chung-Ju Wu <jasonwucj@gmail.com>
23276 * config/nds32/constants.md (unspec_element): New enum.
23277 * config/nds32/constraints.md (Umw): New constraint.
23278 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
23279 * config/nds32/nds32-intrinsic.md: Likewise.
23280 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
23281 (nds32_valid_smw_lwm_base_p): New.
23282 (nds32_output_smw_single_word): New.
23283 (nds32_output_lmw_single_word): New.
23284 (nds32_expand_unaligned_load): New.
23285 (nds32_expand_unaligned_store): New.
23286 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
23287 (nds32_output_smw_single_word): Declare.
23288 (nds32_output_lmw_single_word): Declare.
23289 (nds32_expand_unaligned_load): Declare.
23290 (nds32_expand_unaligned_store): Declare.
23291 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
23292 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
23293 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
23294 NDS32_BUILTIN_UASTORE_DW.
23295 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
23298 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
23299 Kito Cheng <kito.cheng@gmail.com>
23300 Chung-Ju Wu <jasonwucj@gmail.com>
23302 * config/nds32/nds32-intrinsic.c
23303 (nds32_expand_builtin_null_ftype_reg): Delete.
23304 (nds32_expand_builtin_reg_ftype_imm): Ditto.
23305 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
23306 (nds32_read_argument): New.
23307 (nds32_legitimize_target): Ditto.
23308 (nds32_legitimize_argument): Ditto.
23309 (nds32_check_constant_argument): Ditto.
23310 (nds32_expand_unop_builtin): Ditto.
23311 (nds32_expand_unopimm_builtin): Ditto.
23312 (nds32_expand_binop_builtin): Ditto.
23313 (nds32_builtin_decl_impl): Ditto.
23314 (builtin_description): Ditto.
23315 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
23316 (nds32_init_builtins_impl): Ditto.
23317 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
23318 (nds32_builtin_decl): New.
23319 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
23320 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
23322 2018-03-02 Jeff Law <law@redhat.com>
23324 * reorg.c (stop_search_p): Handle DEBUG_INSN.
23325 (redundant_insn, fill_simple_delay_slots): Likewise.
23326 (fill_slots_from_thread): Likewise.
23327 * resource.c (mark_referenced_resources): Likewise.
23328 (mark_set_resources, find_dead_or_set_registers): Likewise.
23330 2018-03-02 Jakub Jelinek <jakub@redhat.com>
23332 * substring-locations.h (format_warning_va): Formatting fix for
23333 ATTRIBUTE_GCC_DIAG.
23334 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
23336 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
23337 * substring-locations.c: Include intl.h.
23338 (format_warning_va): Turned into small wrapper around
23339 format_warning_n_va, renamed to ...
23340 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
23341 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
23343 (format_warning_at_substring_n): New function.
23344 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
23345 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
23346 format_warning_at_substring with just a shorter name instead of
23347 const function pointer.
23348 (fmtwarn_n): New function.
23349 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
23350 appropriate, get rid of all the fmtstr temporaries, move conditionals
23351 with G_() wrapped string literals directly into fmtwarn arguments,
23352 cast dir.len to (int), formatting fixes.
23354 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
23356 * doc/invoke.texi: Remove "Cilk Plus" references.
23358 2018-03-02 Jakub Jelinek <jakub@redhat.com>
23359 Richard Biener <rguenther@suse.de>
23362 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
23363 for error or warning attributes if CALL_FROM_THUNK_P is set.
23366 2018-03-02 Jakub Jelinek <jakub@redhat.com>
23369 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
23370 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
23373 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
23374 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
23376 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
23377 instead of -1U in last predictors element's probability member.
23379 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
23382 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
23383 arguments if they are comparable.
23385 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
23387 PR tree-optimization/84634
23388 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
23389 masks and masked_loop_p with a single loop_masks, making sure it's
23390 null for bb vectorization.
23392 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
23394 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
23395 (vect_analyze_data_ref_access): Use loop->safe_len rather than
23396 loop->force_vectorize to check whether there is no alias.
23398 2018-03-02 Jakub Jelinek <jakub@redhat.com>
23401 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
23403 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
23405 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
23406 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
23407 instead of a loop around prev_real_insn.
23408 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
23411 PR inline-asm/84625
23412 * config/i386/i386.c (ix86_print_operand): Use conditional
23413 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
23416 2018-03-02 Richard Biener <rguenther@suse.de>
23418 PR tree-optimization/84427
23419 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
23420 (bitmap_set_subtract_values): Rewrite to handle multiple
23423 (prune_clobbered_mems): Likewise.
23424 (phi_translate): Take edge instead of pred/phiblock.
23425 (phi_translate_1): Likewise.
23426 (phi_translate_set): Likewise. Insert all translated
23427 exprs for a value into the set, keeping possibly multiple
23428 expressions per value.
23429 (compute_antic_aux): Adjust for phi_translate changes.
23430 When intersecting union the expressions and prune those
23431 not in the final value set, keeping possibly multiple
23432 expressions per value. Do not use value-insertion
23433 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
23434 all expressions. Add verification that the value-sets
23435 only shrink during iteration.
23436 (compute_partial_antic_aux): Adjust for the phi_translate changes.
23437 (do_pre_regular_insertion): Likewise.
23438 (do_pre_partial_partial_insertion): Likewise.
23440 2018-03-02 Richard Biener <rguenther@suse.de>
23443 * config/darwin.c (saved_debug_info_level): New static global.
23444 (darwin_asm_lto_start): Disable debug info generation for LTO out.
23445 (darwin_asm_lto_end): Restore debug info generation settings.
23447 2018-03-01 Martin Liska <mliska@suse.cz>
23450 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
23451 volatile arguments.
23453 2018-03-01 Richard Biener <rguenther@suse.de>
23456 * dwarf2out.c (gen_variable_die): Properly handle late VLA
23457 type annotation with LTO when debug was disabled at compile-time.
23459 2018-03-01 Matthew Fortune <mfortune@gmail.com>
23461 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
23463 (mips_final_postscan_insn): Likewise.
23465 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
23467 PR rtl-optimization/84528
23468 * alias.c (init_alias_target): Add commentary.
23469 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
23470 a unique base value if the frame pointer is not eliminated
23471 to the stack pointer.
23473 2018-03-01 Tom de Vries <tom@codesourcery.com>
23475 PR rtl-optimization/83327
23476 * lra-int.h (hard_regs_spilled_into): Declare.
23477 * lra.c (hard_regs_spilled_into): Define.
23478 (init_reg_info): Init hard_regs_spilled_into.
23479 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
23480 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
23481 (process_bb_lives): Handle hard_regs_spilled_into.
23482 (lra_create_live_ranges_1): Before doing liveness propagation, clear
23483 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
23485 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
23487 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
23488 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
23489 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
23490 * config/rs6000/aix72.h: New file.
23492 2018-02-28 Jakub Jelinek <jakub@redhat.com>
23494 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
23495 instead of warning_at with conditional singular and plural messages
23499 * stor-layout.c (update_alignment_for_field): For
23500 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
23501 && !DECL_PACKED (field), do the alignment update, just use
23502 only desired_align instead of MAX (type_align, desired_align)
23504 (place_field): Don't do known_align < desired_align handling
23505 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
23506 is non-NULL, instead do it after rli->prev_field handling and
23507 only if not within a bitfield word. For DECL_PACKED (field)
23508 use type_align of BITS_PER_UNIT.
23510 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
23512 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
23513 superfluous parentheses and trailing spaces.
23515 2018-02-28 Richard Biener <rguenther@suse.de>
23517 PR tree-optimization/84584
23518 * graphite-scop-detection.c (scop_detection::add_scop): Discard
23519 SCoPs with fake exit edge.
23521 2018-02-28 Martin Liska <mliska@suse.cz>
23524 * timevar.c (timer::print): Fix format to properly print 100%
23527 2018-02-28 Richard Biener <rguenther@suse.de>
23529 PR middle-end/84607
23530 * genmatch.c (capture_info::walk_match): Do not mark
23531 captured expressions without operands as expr_p given
23532 they act more like predicates and should be subject to
23533 "lost tail" side-effect preserving.
23535 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
23537 PR rtl-optimization/81611
23538 * auto-inc-dec.c (attempt_change): Move dead note from
23539 mem_insn if it's the next use of regno
23540 (find_address): Take address use of reg holding
23541 non-incremented value. Add parm to limit search to the named
23543 (merge_in_block): Attempt to use a mem insn that is the next
23544 use of the original regno.
23546 2018-02-27 Martin Sebor <msebor@redhat.com>
23549 * doc/invoke.texi (-Wmissing-attributes): New option.
23550 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
23552 2018-02-27 Martin Sebor <msebor@redhat.com>
23554 PR translation/84207
23555 * diagnostic-core.h (warning_n, error_n, inform_n): Change
23556 n argument to unsigned HOST_WIDE_INT.
23557 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
23558 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
23559 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
23560 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
23562 2018-02-27 Richard Biener <rguenther@suse.de>
23564 PR tree-optimization/84512
23565 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
23566 Do not use the estimate returned from record_stmt_cost for
23567 the scalar iteration cost but sum properly using add_stmt_cost.
23569 2018-02-27 Richard Biener <rguenther@suse.de>
23571 PR tree-optimization/84466
23572 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
23573 Adjust last change to less strictly validate use operands.
23575 2018-02-27 Martin Liska <mliska@suse.cz>
23577 PR gcov-profile/84548
23578 * gcov.c (process_file): Allow partial overlap and consider it
23579 also as group functions.
23580 (output_lines): Properly calculate range of lines for a group.
23582 2018-02-27 Martin Liska <mliska@suse.cz>
23584 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
23585 'ggc' suffixes. Change first column width.
23586 (timer::print): Fix formatting of the column.
23588 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
23590 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
23591 preserve inline entry blocks for the sake of debug inline
23592 entry point markers alone.
23593 (remove_unused_locals): Suggest in comments a better place to
23594 force the preservation of inline entry blocks that are
23595 otherwise unused, but do not preserve them.
23597 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
23599 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
23601 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
23604 * config/i386/constraints.md (Bs): Replace
23605 ix86_indirect_branch_register with
23606 TARGET_INDIRECT_BRANCH_REGISTER.
23608 * config/i386/i386.md (indirect_jump): Likewise.
23609 (tablejump): Likewise.
23610 (*sibcall_memory): Likewise.
23611 (*sibcall_value_memory): Likewise.
23612 Peepholes of indirect call and jump via memory: Likewise.
23613 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
23614 (*sibcall_value_GOT_32): Likewise.
23615 * config/i386/predicates.md (indirect_branch_operand): Likewise.
23616 (GOT_memory_operand): Likewise.
23617 (call_insn_operand): Likewise.
23618 (sibcall_insn_operand): Likewise.
23619 (GOT32_symbol_operand): Likewise.
23620 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
23622 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
23624 PR rtl-optimization/83496
23625 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
23626 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
23627 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
23628 redundant insn, if any.
23629 (relax_delay_slots): Likewise.
23630 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
23632 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
23634 PR tree-optimization/83965
23635 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
23636 that grouped statements are part of a reduction chain. Return
23637 true if the statement is not marked as a reduction itself but
23638 is part of a group.
23639 (vect_recog_dot_prod_pattern): Don't check whether the statement
23640 is part of a group here.
23641 (vect_recog_sad_pattern): Likewise.
23642 (vect_recog_widen_sum_pattern): Likewise.
23644 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
23647 * final.c (rest_of_clean_state): Also look for calls inside sequences.
23649 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
23652 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
23654 (ix86_output_indirect_function_return): New prototype.
23655 (ix86_split_simple_return_pop_internal): Likewise.
23656 * config/i386/i386.c (indirect_return_via_cx): New.
23657 (indirect_return_via_cx_bnd): Likewise.
23658 (indirect_thunk_name): Handle return va CX_REG.
23659 (output_indirect_thunk_function): Create alias for
23660 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
23661 (ix86_output_indirect_jmp): Remove the bool argument.
23662 (ix86_output_indirect_function_return): New function.
23663 (ix86_split_simple_return_pop_internal): Likewise.
23664 * config/i386/i386.md (*indirect_jump): Don't pass false
23665 to ix86_output_indirect_jmp.
23666 (*tablejump_1): Likewise.
23667 (simple_return_pop_internal): Change it to define_insn_and_split.
23668 Call ix86_split_simple_return_pop_internal to split it for
23669 -mfunction-return=.
23670 (simple_return_indirect_internal): Call
23671 ix86_output_indirect_function_return instead of
23672 ix86_output_indirect_jmp.
23674 2018-02-26 Jakub Jelinek <jakub@redhat.com>
23677 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
23678 memset and value initialization afterwards.
23680 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
23682 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
23684 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
23687 * common/config/aarch64/aarch64-common.c
23688 (aarch_option_optimization_table[]): Switch
23689 off fomit-frame-pointer
23691 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
23692 Chung-Ju Wu <jasonwucj@gmail.com>
23694 * config/nds32/nds32-multiple.md (load_multiple): Disallow
23696 (store_multiple): Ditto.
23698 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
23700 * config.gcc: Add --with-cpu support for nds32 target.
23701 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
23702 * config/nds32/nds32.opt: Add -mcpu= option.
23704 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
23706 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
23707 isel=yes): Warn for these deprecated options.
23709 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
23711 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
23712 ISA_2_5_MASKS_EMBEDDED.
23714 2018-02-23 Jakub Jelinek <jakub@redhat.com>
23716 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
23717 p->max as pointers rather than using iterative_hash_expr.
23719 2018-02-23 Carl Love <cel@us.ibm.com>
23721 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
23722 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
23724 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
23725 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
23726 P8V_BUILTIN_VEC_VUNSIGNED2.
23728 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
23731 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
23732 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
23734 * lra-constraints.c (curr_insn_transform): Set up
23735 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
23737 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
23738 * lra-eliminations.c (spill_pseudos): Ditto.
23739 (process_insn_for_elimination): Ditto.
23740 * lra-lives.c (reg_early_clobber_p): Use the new macros.
23741 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
23742 LRA_NON_CLOBBERED_ALT.
23744 2018-02-22 Martin Sebor <msebor@redhat.com>
23746 PR tree-optimization/84480
23747 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
23748 to maybe_diag_stxncpy_trunc. Call it.
23749 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
23750 from gimple_fold_builtin_strcpy. Print inlining stack.
23751 (handle_builtin_stxncpy): Print inlining stack.
23752 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
23754 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
23757 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
23758 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
23759 and -fcheck-pointer-bounds are used together.
23760 (indirect_thunk_prefix): New enum.
23761 (indirect_thunk_need_prefix): New function.
23762 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
23763 "_nt" instead of "_bnd" for NOTRACK prefix.
23764 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
23765 (output_indirect_thunk_function): Likewise.
23767 (ix86_code_end): Update output_indirect_thunk_function calls.
23768 (ix86_output_indirect_branch_via_reg): Replace
23769 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
23770 (ix86_output_indirect_branch_via_push): Likewise.
23771 (ix86_output_function_return): Likewise.
23772 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
23773 incompatible with -fcf-protection=branch and
23774 -fcheck-pointer-bounds.
23776 2018-02-22 Steve Ellcey <sellcey@cavium.com>
23779 * config/aarch64/aarch64.c (aarch64_print_address_internal):
23780 Change gcc_assert call to output_operand_lossage.
23782 2018-02-22 Steve Ellcey <sellcey@cavium.com>
23784 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
23786 2018-02-22 DJ Delorie <dj@redhat.com>
23787 Sebastian Perta <sebastian.perta@renesas.com>
23788 Oleg Endo <olegendo@gcc.gnu.org>
23790 * config/rx/rx.c (rx_rtx_costs): New function.
23791 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
23793 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
23795 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
23797 2018-02-22 Martin Liska <mliska@suse.cz>
23800 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
23801 Add "native" as a possible value.
23803 2018-02-22 Martin Liska <mliska@suse.cz>
23806 * config/i386/i386.c (ix86_option_override_internal):
23807 Add "native" as a possible value for -march and -mtune.
23809 2018-02-22 Jakub Jelinek <jakub@redhat.com>
23812 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
23813 to all type variants.
23815 PR tree-optimization/84503
23816 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
23817 width as info->bitpos + info->bitsize - start.
23818 (merged_store_group::merge_overlapping): Simplify width computation.
23819 (check_no_overlap): New function.
23820 (imm_store_chain_info::try_coalesce_bswap): Compute expected
23821 start + width and last_order of the group, fail if check_no_overlap
23823 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
23824 to group if check_no_overlap fails.
23826 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
23828 * config/rs6000/altivec.md: Delete contraint arguments to
23829 define_expand, define_split, and define_peephole2, and in
23830 define_insn_and_split if always unused.
23831 * config/rs6000/darwin.md: Ditto.
23832 * config/rs6000/dfp.md: Ditto.
23833 * config/rs6000/rs6000.md: Ditto.
23834 * config/rs6000/sync.md: Ditto.
23835 * config/rs6000/vector.md: Ditto.
23836 * config/rs6000/vsx.md: Ditto.
23838 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
23840 * config/rs6000/altivec.md: Write output control strings as braced
23841 blocks instead of double-quoted strings.
23842 * config/rs6000/darwin.md: Ditto.
23843 * config/rs6000/rs6000.md: Ditto.
23844 * config/rs6000/vector.md: Ditto.
23845 * config/rs6000/vsx.md: Ditto.
23847 2018-02-21 Jason Merrill <jason@redhat.com>
23849 PR c++/84314 - ICE with templates and fastcall attribute.
23850 * attribs.c (build_type_attribute_qual_variant): Remove assert.
23852 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
23854 * ipa-cp.c (determine_versionability): Fix comment typos.
23856 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
23859 * ipa-cp.c (determine_versionability): Do not version functions caling
23862 2018-02-21 Martin Liska <mliska@suse.cz>
23865 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
23866 Add "native" as a possible value.
23867 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
23868 the macro when native cpu detection is available.
23870 2018-02-21 Martin Liska <mliska@suse.cz>
23873 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
23874 Add "native" as a possible value.
23875 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
23876 when native cpu detection is available.
23878 2018-02-21 Jakub Jelinek <jakub@redhat.com>
23879 Martin Sebor <msebor@redhat.com>
23881 PR tree-optimization/84478
23882 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
23884 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
23885 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
23886 support which is conservatively correct, for 2 only stay conservative
23887 for maxlen. Formatting and comment capitalization fixes. Add STRICT
23888 argument to the 2 argument get_range_strlen, adjust 6 arg
23889 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
23891 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
23892 (gimple_fold_builtin_strlen): Pass true as last argument to
23895 2018-02-20 Martin Sebor <msebor@redhat.com>
23897 PR middle-end/84095
23898 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
23899 (builtin_memref::set_base_and_offset): Same. Handle inner references.
23900 (builtin_memref::builtin_memref): Factor out parts into
23901 set_base_and_offset and call it.
23903 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
23905 PR middle-end/84406
23906 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
23907 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
23908 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
23909 search at the associated MODE_INT.
23911 2018-02-20 Jeff Law <law@redhat.com>
23913 PR middle-end/82123
23914 PR tree-optimization/81592
23915 PR middle-end/79257
23916 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
23917 for range data rather than using global data.
23918 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
23919 range data rather than using global data.
23920 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
23921 pass it to children as needed.
23922 (struct directive::fmtresult): Similarly.
23923 (struct directive::set_width): Similarly.
23924 (struct directive::set_precision): Similarly.
23925 (format_integer, format_directive, parse_directive): Similarly.
23926 (format_none): Accept unnamed vr_values parameter.
23927 (format_percent, format_floating, format_character): Similarly.
23928 (format_string, format_plain): Similarly.
23929 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
23930 the EVRP range analyzer for range data rather than using global data.
23931 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
23932 gimple-ssa-evrp-analyze.h
23933 (class sprintf_dom_walker): Add after_dom_children member function.
23934 Add evrp_range_analyzer member.
23935 (sprintf_dom_walker::before_dom_children): Call into the EVRP
23936 range analyzer as needed.
23937 (sprintf_dom_walker::after_dom_children): New member function.
23938 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
23940 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
23941 (evrp_range_analyzer::pop_to_marker): Likewise.
23943 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
23945 PR tree-optimization/84419
23946 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
23947 with the required type if its current type is compatible but
23950 2018-02-20 Jakub Jelinek <jakub@redhat.com>
23952 PR middle-end/82004
23953 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
23954 after vectorization.
23956 2018-02-20 Martin Liska <mliska@suse.cz>
23959 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
23960 possible values if we don't have a hint.
23962 2018-02-20 Martin Liska <mliska@suse.cz>
23966 * final.c (shorten_branches): Build align_tab array with one
23968 * opts.c (finish_options): Add alignment option limit check.
23969 (MAX_CODE_ALIGN): Likewise.
23970 (MAX_CODE_ALIGN_VALUE): Likewise.
23971 * doc/invoke.texi: Document maximum allowed option value for
23972 all -falign-* options.
23974 2018-02-19 Jakub Jelinek <jakub@redhat.com>
23977 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
23978 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
23979 * var-tracking.c (emit_note_insn_var_location): Remove all references
23980 to NOTE_INSN_CALL_ARG_LOCATION.
23981 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
23982 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
23983 Use copy_rtx_if_shared.
23984 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
23985 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
23986 (dwarf2out_var_location): Remove handling of
23987 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
23989 * final.c (final_scan_insn): Remove all references to
23990 NOTE_INSN_CALL_ARG_LOCATION.
23991 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
23992 before dumping final insns.
23993 * except.c (emit_note_eh_region_end): Remove all references to
23994 NOTE_INSN_CALL_ARG_LOCATION.
23995 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
23996 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
23997 * config/arc/arc.c (hwloop_optimize): Likewise.
23998 * config/arm/arm.c (create_fix_barrier): Likewise.
23999 * config/s390/s390.c (s390_chunkify_start): Likewise.
24000 * config/sh/sh.c (find_barrier): Likewise.
24001 * config/i386/i386.c (rest_of_insert_endbranch,
24002 ix86_seh_fixup_eh_fallthru): Likewise.
24003 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
24004 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
24005 * config/frv/frv.c (frv_function_prologue): Likewise.
24006 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
24008 (note_outside_basic_block_p): Remove all references to
24009 NOTE_INSN_CALL_ARG_LOCATION.
24010 * gengtype.c (adjust_field_rtx_def): Likewise.
24011 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
24013 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
24014 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
24017 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
24020 PR tree-optimization/84452
24021 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
24022 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
24025 2018-02-19 Martin Liska <mliska@suse.cz>
24028 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
24030 2018-02-19 Martin Liska <mliska@suse.cz>
24031 Richard Sandiford <richard.sandiford@linaro.org>
24033 PR tree-optimization/82491
24034 * gimple-fold.c (get_base_constructor): Make earlier bail out
24037 2018-02-19 Carl Love <cel@us.ibm.com>
24039 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
24040 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
24042 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
24043 P8V_BUILTIN_VEC_NEG.
24045 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
24047 * config/rl78/rl78.md (movdf): New define expand.
24049 2018-02-19 Martin Liska <mliska@suse.cz>
24052 * doc/invoke.texi: Fix typo.
24053 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
24055 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
24057 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
24058 handle rs6000_single_float and rs6000_double_float specially for
24061 2018-02-16 Jeff Law <law@redhat.com>
24063 * config/rx/rx.c (add_pop_cfi_notes): New function.;
24064 (pop_regs): Use it.
24066 2018-02-16 Jakub Jelinek <jakub@redhat.com>
24069 * ipa-inline.c (inline_small_functions): Fix a typo.
24071 2018-02-16 Nathan Sidwell <nathan@acm.org>
24073 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
24075 2018-02-16 Carl Love <cel@us.ibm.com>
24077 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
24078 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
24079 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
24080 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
24081 expansion to P8V_BUILTIN_VEC_FLOAT2.
24083 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
24085 PR rtl-optimization/70023
24086 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
24087 src_regno into account.
24089 2018-02-16 Carl Love <cel@us.ibm.com>
24091 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
24092 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
24093 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
24094 * config/rs6000/rs6000.c: Remove case statements for
24095 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
24096 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
24097 and P9V_BUILTIN_VEC_VINSERT4B.
24098 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
24099 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
24100 * config/rs6000/vsx.md:
24101 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
24104 2018-02-16 Carl Love <cel@us.ibm.com>
24106 * config/rs6000/altivec.h: Add builtin names vec_extract4b
24108 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
24110 * config/rs6000/rs6000-c.c: Add the definitions for
24111 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
24112 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
24113 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
24114 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
24115 definition for insert4b and define insn *insert3b_internal.
24116 * doc/extend.texi: Add documentation for vec_extract4b.
24118 2018-02-16 Nathan Sidwell <nathan@acm.org>
24120 * doc/extend.texi (Backwards Compatibility): Mention friend
24121 injection. Note for-scope is deprecated.
24122 * doc/invoke.texi (-ffriend-injection): Deprecate.
24124 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
24126 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
24127 that moved to I2, also allow destinations that are a paradoxical
24128 subreg (instead of a normal reg).
24130 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
24133 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
24136 2018-02-16 Richard Biener <rguenther@suse.de>
24138 PR tree-optimization/84037
24139 PR tree-optimization/84016
24141 * config/i386/i386.c (ix86_builtin_vectorization_cost):
24142 Adjust vec_construct for the fact we need additional higher latency
24143 128bit inserts for AVX256 and AVX512 vector builds.
24144 (ix86_add_stmt_cost): Scale vector construction cost for
24147 2018-02-16 Richard Biener <rguenther@suse.de>
24149 PR tree-optimization/84417
24150 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
24151 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
24152 (non_rewritable_lvalue_p): Likewise, use poly-ints.
24154 2018-02-16 Martin Liska <mliska@suse.cz>
24157 * internal-fn.def (ASAN_CHECK): Set proper flags.
24158 (ASAN_MARK): Likewise.
24160 2018-02-16 Julia Koval <julia.koval@intel.com>
24162 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
24163 from PTA_CANNONLAKE.
24165 2018-02-16 Jakub Jelinek <jakub@redhat.com>
24168 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
24169 Use ++iter rather than iter++ for std::list iterators.
24170 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
24171 defer deleting them until all nodes in the forest are processed. Do
24172 free even leaf nodes. Change to_process into auto_vec.
24175 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
24176 * vec.h (vec_default_construct): Use memset instead of placement new
24177 if BROKEN_VALUE_INITIALIZATION is defined.
24178 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
24179 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
24182 PR rtl-optimization/83723
24183 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
24184 * lra.c (lra_substitute_pseudo): Likewise. If true, use
24185 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
24187 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
24189 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
24191 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
24193 PR rtl-optimization/81443
24194 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
24195 from inner REGs to paradoxical SUBREGs.
24197 2018-02-16 Richard Biener <rguenther@suse.de>
24199 PR tree-optimization/84399
24200 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
24201 For operands we can analyze at their definition make sure we can
24202 analyze them at each use as well.
24204 2018-02-16 Richard Biener <rguenther@suse.de>
24206 PR tree-optimization/84190
24207 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
24208 volatile accesses if the decl isn't volatile.
24210 2018-02-15 Jason Merrill <jason@redhat.com>
24212 PR c++/84314 - ICE with templates and fastcall attribute.
24213 * attribs.c (build_type_attribute_qual_variant): Don't clobber
24214 TYPE_CANONICAL on an existing type.
24216 2018-02-15 Jakub Jelinek <jakub@redhat.com>
24218 PR tree-optimization/84383
24219 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
24220 dstoff nor call operand_equal_p if dstbase is NULL.
24222 PR tree-optimization/84334
24223 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
24224 also a CONSTANT_CLASS_P, punt.
24226 2018-02-14 Jim Wilson <jimw@sifive.com>
24228 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
24229 first SMALL_OPERAND check. New local min_second_step. Move assert
24230 to where locals are set. Add TARGET_RVC support.
24231 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
24233 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
24235 * doc/invoke.texi: Correct -Wformat-overflow code sample.
24237 2018-02-14 Martin Sebor <msebor@redhat.com>
24239 PR tree-optimization/83698
24240 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
24241 arrays constrain the offset range to their bounds.
24242 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
24243 (builtin_access::overlap): Avoid setting the size of overlap if it's
24245 (maybe_diag_overlap): Also consider arrays when deciding what values
24246 of offsets to include in diagnostics.
24248 2018-02-14 Martin Sebor <msebor@redhat.com>
24251 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
24252 that correspond to the kind of a declaration.
24254 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
24257 * config/pa/pa.md: Load address of PIC label using the linkage table
24258 if the label is nonlocal.
24260 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
24262 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
24263 warning message if user requests -maltivec=be.
24264 * doc/invoke.texi: Document deprecation of -maltivec=be.
24266 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
24269 * config/rs6000/rs6000-c.c: Update definitions for
24270 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
24271 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
24273 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
24276 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
24277 add _get_ssp intrinsics. Remove argument from
24278 __builtin_ia32_rdssp[d|q].
24279 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
24280 * config/i386/i386-builtin.def: Remove argument from
24281 __builtin_ia32_rdssp[d|q].
24282 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
24283 ix86_expand_special_args_builtin for _rdssp[d|q].
24284 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
24285 Clear register before usage.
24286 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
24287 Add documentation for new _get_ssp and _inc_ssp intrinsics.
24289 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
24291 PR tree-optimization/84357
24292 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
24293 operand 1 of an ARRAY_REF too.
24295 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
24298 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
24299 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
24301 (set_of_reg): New struct.
24302 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
24303 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
24304 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
24306 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
24307 Split into bitclr, bitset, bitinvert patterns if appropriate.
24308 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
24309 use rx_fuse_in_memory_bitop.
24310 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
24311 to named insn, correct maximum insn length.
24313 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
24316 * machmode.def: Define a complex mode for PARTIAL_INT.
24317 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
24319 * doc/rtl.texi: Document CSPImode.
24320 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
24322 (msp430_hard_regno_nregs_with_padding): Likewise.
24324 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
24327 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
24329 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
24331 PR rtl-optimization/84169
24332 * combine.c (try_combine): New variable split_i2i3. Set it to true if
24333 we generated a parallel as new i3 and we split that to new i2 and i3
24334 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
24335 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
24336 those to i2, not i1. Partially rewrite this scan code.
24338 2018-02-13 Jakub Jelinek <jakub@redhat.com>
24341 * stor-layout.c (place_field): For variable length fields, adjust
24342 offset_align afterwards not just based on the field's alignment,
24343 but also on the size.
24345 PR middle-end/84309
24346 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
24347 of exps and logs in the use_exp2 case.
24349 2018-02-13 Jeff Law <law@redhat.com>
24351 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
24352 entry for "vector".
24354 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
24357 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
24361 * common.opt (gas-loc-support, gas-locview-support): New.
24362 (ginline-points, ginternal-reset-location-views): New.
24363 * doc/invoke.texi: Document them. Use @itemx where intended.
24364 (gvariable-location-views): Adjust.
24365 * target.def (reset_location_view): New.
24366 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
24367 (TARGET_RESET_LOCATION_VIEW): New.
24368 * doc/tm.texi: Rebuilt.
24369 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
24370 (dwarf2out_default_as_locview_support): New.
24371 (output_asm_line_debug_info): Use option variables.
24372 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
24373 (output_loc_list): Likewise.
24374 (add_high_low_attributes): Check option variables.
24375 Don't output entry view attribute in strict mode.
24376 (gen_inlined_subroutine_die): Check option variables.
24377 (dwarf2out_inline_entry): Likewise.
24378 (init_sections_and_labels): Likewise.
24379 (dwarf2out_early_finish): Likewise.
24380 (maybe_reset_location_view): New, from...
24381 (dwarf2out_var_location): ... here. Call it.
24382 * debug.h (dwarf2out_default_as_loc_support): Declare.
24383 (dwarf2out_default_as_locview_support): Declare.
24384 * hooks.c (hook_int_rtx_insn_0): New.
24385 * hooks.h (hook_int_rtx_insn_0): Declare.
24386 * toplev.c (process_options): Take -gas-loc-support and
24387 -gas-locview-support from dwarf2out. Enable
24388 -gvariable-location-views by default only with locview
24389 assembler support. Enable -ginternal-reset-location-views by
24390 default only if the target defines the corresponding hook.
24391 Enable -ginline-points by default if location views are
24392 enabled; force it disabled if statement frontiers are
24394 * tree-inline.c (expand_call_inline): Check option variables.
24395 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
24397 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
24399 PR tree-optimization/84321
24400 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
24401 handling. Also check whether the anti-range contains any values
24402 that satisfy the mask; switch to a VR_RANGE if not.
24404 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
24407 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
24409 2018-02-13 Martin Jambor <mjambor@suse.cz>
24412 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
24413 of call statements, also set location of a load to a temporary.
24415 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
24417 * config/rl78/rl78.c (add_vector_labels): New function.
24418 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
24419 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
24420 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
24421 which checks that no arguments are passed.
24422 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
24423 * doc/extend.texi: Documentation for the new attribute.
24425 2018-02-13 Andreas Schwab <schwab@suse.de>
24427 * config/riscv/linux.h (CPP_SPEC): Define.
24429 2018-02-13 Jakub Jelinek <jakub@redhat.com>
24432 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
24433 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
24434 OPTION_MASK_ISA_AES as first argument to def_builtin_const
24435 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
24436 instead of OPTION_MASK_ISA_PCLMUL as first argument to
24437 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
24438 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
24439 temporarily for AES and PCLMUL builtins.
24441 PR tree-optimization/84339
24442 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
24443 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
24446 PR middle-end/84309
24447 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
24448 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
24449 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
24451 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
24453 * omp-simd-clone.h: New file.
24454 * omp-simd-clone.c: Include omp-simd-clone.h.
24455 (expand_simd_clones): No longer static.
24456 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
24457 cgraph.h and omp-simd-clone.h.
24458 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
24459 (vect_recog_widen_shift_pattern): Formatting fix.
24460 (vect_pattern_recog_1): Don't check optab for calls.
24463 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
24464 operands[2] into a REG before using gen_lowpart on it.
24466 2018-02-12 Jeff Law <law@redhat.com>
24469 * config/sh/sh.c (find_barrier): Consider a sibling call
24472 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
24473 successfully back substituting a reg.
24475 2018-02-12 Richard Biener <rguenther@suse.de>
24477 PR tree-optimization/84037
24478 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
24479 parameter, move visited init to caller.
24480 (vect_slp_analyze_operations): Separate cost from validity
24481 check, initialize visited once for all instances.
24482 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
24484 * tree-vect-stmts.c (vect_model_simple_cost): Make early
24486 (vect_model_promotion_demotion_cost): Likewise.
24487 (vectorizable_bswap): Guard cost modeling with !slp_node
24488 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
24490 (vectorizable_call): Likewise.
24491 (vectorizable_conversion): Likewise.
24492 (vectorizable_assignment): Likewise.
24493 (vectorizable_shift): Likewise.
24494 (vectorizable_operation): Likewise.
24495 (vectorizable_store): Likewise.
24496 (vectorizable_load): Likewise.
24497 (vectorizable_condition): Likewise.
24498 (vectorizable_comparison): Likewise.
24500 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
24503 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
24504 (ASAN_MARK): Fix fnspec to account for return value, change pointer
24505 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
24507 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
24509 PR middle-end/83665
24510 * params.def (inline-min-speedup): Increase from 8 to 15.
24511 (max-inline-insns-auto): Decrease from 40 to 30.
24512 * ipa-split.c (consider_split): Add some buffer for function to
24513 be considered inlining candidate.
24514 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
24517 2018-02-12 Richard Biener <rguenther@suse.de>
24519 PR tree-optimization/84037
24520 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
24521 matched stmts if we cannot swap the non-matched ones.
24523 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
24525 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
24526 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
24527 _mm_maskz_scalef_round_ss): New intrinsics.
24528 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
24529 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
24530 __builtin_ia32_scalefss_round): Remove.
24531 (__builtin_ia32_scalefsd_mask_round,
24532 __builtin_ia32_scalefss_mask_round): New intrinsics.
24533 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
24534 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
24535 ((match_operand:VF_128 2 "<round_nimm_predicate>"
24536 "<round_constraint>")): Changed to ...
24537 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
24538 "<round_scalar_constraint>")): ... this.
24539 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
24540 %0, %1, %2<round_op3>}"): Changed to ...
24541 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
24542 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
24543 %2<round_scalar_mask_op3>}"): ... this.
24544 * config/i386/subst.md (round_scalar_nimm_predicate): New.
24546 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
24548 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
24549 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
24550 (_mm_maskz_sqrt_round_ss): New intrinsics.
24551 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
24552 (__builtin_ia32_sqrtsd_mask_round)
24553 (__builtin_ia32_sqrtss_mask_round): New builtins.
24554 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
24555 (__builtin_ia32_sqrtss_round): Remove.
24556 (__builtin_ia32_sqrtsd_mask_round)
24557 (__builtin_ia32_sqrtss_mask_round): New builtins.
24558 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
24559 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
24560 ((match_operand:VF_128 1 "vector_operand"
24561 "xBm,<round_constraint>")): Changed to ...
24562 ((match_operand:VF_128 1 "vector_operand"
24563 "xBm,<round_scalar_constraint>")): ... this.
24564 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
24565 %0, %2, %<iptr>1<round_op3>}): Changed to ...
24566 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
24567 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
24568 %<iptr>1<round_scalar_mask_op3>}): ... this.
24569 ((set_attr "prefix" "<round_prefix>")): Changed to ...
24570 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
24572 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
24575 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
24576 Cast vec_cmpeq result to correct type.
24577 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
24578 Cast vec_cmpgt result to correct type.
24580 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
24582 * final.c (final_scan_insn_1): Renamed from...
24583 (final_scan_insn): ... this. New wrapper, to recover
24584 seen from the outermost call in recursive ones.
24585 * config/sparc/sparc.c (output_return): Drop seen from call.
24586 (output_sibcall): Likewise.
24587 * config/visium/visium.c (output_branch): Likewise.
24589 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
24591 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
24594 2018-02-10 Alan Modra <amodra@gmail.com>
24597 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
24598 Specify LR as an input.
24600 2018-02-10 Jakub Jelinek <jakub@redhat.com>
24603 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
24604 remove_member_access_dummy_vars): New functions.
24605 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
24606 lower_omp_1, execute_lower_omp): Use them.
24608 PR rtl-optimization/84308
24609 * shrink-wrap.c (spread_components): Release todo vector.
24611 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
24613 PR rtl-optimization/57193
24614 * ira-color.c (struct allocno_color_data): Add member
24615 conflict_allocno_hard_prefs.
24616 (update_conflict_allocno_hard_prefs): New.
24617 (bucket_allocno_compare_func): Add a preference based on
24618 conflict_allocno_hard_prefs.
24619 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
24620 (color_allocnos): Remove a dead code. Initiate
24621 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
24623 2018-02-09 Jakub Jelinek <jakub@redhat.com>
24626 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
24627 constraint from =wa to wa. Avoid a subreg on the output operand,
24628 instead use a pseudo and subreg it in a move.
24629 (p9_xxbrd_<mode>): Changed to ...
24630 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
24631 (p9_xxbrd_v2df): New expander.
24632 (p9_xxbrw_<mode>): Changed to ...
24633 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
24634 (p9_xxbrw_v4sf): New expander.
24636 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
24638 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
24640 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
24643 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
24644 multiply in 32-bit mode.
24645 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
24646 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
24649 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
24651 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
24652 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
24653 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
24654 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
24656 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
24659 * dwarf2out.c (is_trivial_indirect_ref): New function.
24660 (dwarf2out_late_global_decl): Do not generate a location
24661 attribute for variables that have a non-trivial DECL_VALUE_EXPR
24662 and that are not defined in the current unit.
24664 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
24666 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
24667 instead of a libcall for UNORDERED.
24669 2018-02-09 Tamar Christina <tamar.christina@arm.com>
24672 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
24673 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
24675 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
24678 * config/s390/s390.c (s390_set_current_function): Invoke
24679 s390_indirect_branch_settings also if fndecl didn't change.
24681 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
24683 * config/rs6000/rs6000.md (blockage): Set length to zero.
24685 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
24687 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
24689 2018-02-09 Jakub Jelinek <jakub@redhat.com>
24692 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
24693 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
24697 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
24698 PARALLEL incoming that failed vt_get_decl_and_offset check.
24700 PR middle-end/84237
24701 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
24702 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
24704 (get_variable_section): For decls in named .bss* sections pass true as
24705 second argument to bss_initializer_p.
24707 2018-02-09 Marek Polacek <polacek@redhat.com>
24708 Jakub Jelinek <jakub@redhat.com>
24711 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
24712 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
24713 Sync some changes from cxx_fold_indirect_ref.
24715 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
24717 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
24719 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
24720 (BLOCK_INLINE_ENTRY_LABEL): New.
24721 (dwarf2out_var_location): Disregard inline entry markers.
24722 (inline_entry_data): New struct.
24723 (inline_entry_data_hasher): New hashtable type.
24724 (inline_entry_data_hasher::hash): New.
24725 (inline_entry_data_hasher::equal): New.
24726 (inline_entry_data_table): New variable.
24727 (add_high_low_attributes): Add DW_AT_entry_pc and
24728 DW_AT_GNU_entry_view attributes if a pending entry is found
24729 in inline_entry_data_table. Add old entry_pc attribute only
24730 if debug nonbinding markers are disabled.
24731 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
24732 markers are enabled.
24733 (block_within_block_p, dwarf2out_inline_entry): New.
24734 (dwarf2out_finish): Check that no entries remained in
24735 inline_entry_data_table.
24736 * final.c (reemit_insn_block_notes): Handle inline entry notes.
24737 (final_scan_insn, notice_source_line): Likewise.
24738 (rest_of_clean_state): Skip inline entry markers.
24739 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
24741 * gimple.c (gimple_build_debug_inline_entry): New.
24742 * gimple.h (enum gimple_debug_subcode): Add
24743 GIMPLE_DEBUG_INLINE_ENTRY.
24744 (gimple_build_debug_inline_entry): Declare.
24745 (gimple_debug_inline_entry_p): New.
24746 (gimple_debug_nonbind_marker_p): Adjust.
24747 * insn-notes.def (INLINE_ENTRY): New.
24748 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
24749 inline entry marker notes.
24750 (print_insn): Likewise.
24751 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
24752 (INSN_DEBUG_MARKER_KIND): Likewise.
24753 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
24754 * tree-inline.c (expand_call_inline): Build and insert
24755 debug_inline_entry stmt.
24756 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
24757 inline entry blocks early, if nonbind markers are enabled.
24758 (dump_scope_block): Dump fragment info.
24759 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
24760 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
24761 (gimple_build_debug_inline_entry): New.
24762 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
24763 Enable/disable inline entry points too.
24764 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
24765 (DEBUG_INSN): Describe inline entry markers.
24767 * common.opt (gvariable-location-views): New.
24768 (gvariable-location-views=incompat5): New.
24769 * config.in: Rebuilt.
24770 * configure: Rebuilt.
24771 * configure.ac: Test assembler for view support.
24772 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
24773 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
24774 * dwarf2out.c (var_loc_view): New typedef.
24775 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
24776 (dwarf2out_locviews_in_attribute): New.
24777 (dwarf2out_locviews_in_loclist): New.
24778 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
24779 (enum dw_line_info_opcode): Add LI_adv_address.
24780 (struct dw_line_info_table): Add view.
24781 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
24782 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
24783 (zero_view_p): New variable.
24784 (ZERO_VIEW_P): New macro.
24785 (output_asm_line_debug_info): New.
24786 (struct var_loc_node): Add view.
24787 (add_AT_view_list, AT_loc_list): New.
24788 (add_var_loc_to_decl): Add view param. Test it against last.
24789 (new_loc_list): Add view params. Record them.
24790 (AT_loc_list_ptr): Handle loc and view lists.
24791 (view_list_to_loc_list_val_node): New.
24792 (print_dw_val): Handle dw_val_class_view_list.
24793 (size_of_die): Likewise.
24794 (value_format): Likewise.
24795 (loc_list_has_views): New.
24796 (gen_llsym): Set vl_symbol too.
24797 (maybe_gen_llsym, skip_loc_list_entry): New.
24798 (dwarf2out_maybe_output_loclist_view_pair): New.
24799 (output_loc_list): Output view list or entries too.
24800 (output_view_list_offset): New.
24801 (output_die): Handle dw_val_class_view_list.
24802 (output_dwarf_version): New.
24803 (output_compilation_unit_header): Use it.
24804 (output_skeleton_debug_sections): Likewise.
24805 (output_rnglists, output_line_info): Likewise.
24806 (output_pubnames, output_aranges): Update version comments.
24807 (output_one_line_info_table): Output view numbers in asm comments.
24808 (dw_loc_list): Determine current endview, pass it to new_loc_list.
24809 Call maybe_gen_llsym.
24810 (loc_list_from_tree_1): Adjust.
24811 (add_AT_location_description): Create view list attribute if
24812 needed, check it's absent otherwise.
24813 (convert_cfa_to_fb_loc_list): Adjust.
24814 (maybe_emit_file): Call output_asm_line_debug_info for test.
24815 (dwarf2out_var_location): Reset views as needed. Precompute
24816 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
24817 attribute. Set view.
24818 (new_line_info_table): Reset next view.
24819 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
24820 (dwarf2out_source_line): Likewise. Output view resets and labels to
24821 the assembler, or select appropriate line info opcodes.
24822 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
24823 (optimize_string_length): Catch it. Adjust.
24824 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
24825 dw_val_class_view_list, and remove it if no longer needed.
24826 (hash_loc_list): Hash view numbers.
24827 (loc_list_hasher::equal): Compare them.
24828 (optimize_location_lists): Check whether a view list symbol is
24829 needed, and whether the locview attribute is present, and
24830 whether they match. Remove the locview attribute if no longer
24832 (index_location_lists): Call skip_loc_list_entry for test.
24833 (dwarf2out_finish): Call output_asm_line_debug_info for test.
24834 Use output_dwarf_version.
24835 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
24836 (struct dw_val_node): Add val_view_list.
24837 * final.c (SEEN_NEXT_VIEW): New.
24838 (set_next_view_needed): New.
24839 (clear_next_view_needed): New.
24840 (maybe_output_next_view): New.
24841 (final_start_function): Rename to...
24842 (final_start_function_1): ... this. Take pointer to FIRST,
24843 add SEEN parameter. Emit param bindings in the initial view.
24844 (final_start_function): Reintroduce SEEN-less interface.
24845 (final): Rename to...
24846 (final_1): ... this. Take SEEN parameter. Output final pending
24847 next view at the end.
24848 (final): Reintroduce seen-less interface.
24849 (final_scan_insn): Output pending next view before switching
24850 sections or ending a block. Mark the next view as needed when
24851 outputting variable locations. Notify debug backend of section
24852 changes, and of location view changes.
24853 (rest_of_handle_final): Adjust.
24854 * toplev.c (process_options): Autodetect value for debug variable
24855 location views option. Warn on incompat5 without -gdwarf-5.
24856 * doc/invoke.texi (gvariable-location-views): New.
24857 (gvariable-location-views=incompat5): New.
24858 (gno-variable-location-views): New.
24860 2018-02-08 David Malcolm <dmalcolm@redhat.com>
24862 PR tree-optimization/84136
24863 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
24864 that the result of find_edge is non-NULL.
24866 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
24869 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
24870 storing integer register in SImode. Fix cost of 256 and 512
24871 byte aligned SSE register store.
24873 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
24875 * config/i386/i386.c (ix86_multiplication_cost): Fix
24876 multiplication cost for TARGET_AVX512DQ.
24878 2018-02-08 Marek Polacek <polacek@redhat.com>
24880 PR tree-optimization/84238
24881 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
24884 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
24886 PR tree-optimization/84265
24887 * tree-vect-stmts.c (vectorizable_store): Don't treat
24888 VMAT_CONTIGUOUS accesses as grouped.
24889 (vectorizable_load): Likewise.
24891 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
24893 PR tree-optimization/81635
24894 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
24895 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
24896 (test_round_for_mask): New functions.
24897 (wide_int_cc_tests): Call test_round_for_mask.
24898 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
24899 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
24900 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
24901 range returned by get_range_info.
24903 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
24906 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
24907 * symtab.c: Include builtins.h
24908 (symtab_node::output_to_lto_symbol_table_p): Move here
24909 from lto-streamer-out.c:output_symbol_p.
24910 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
24911 (output_symbol_p): Move all logic to symtab.c
24912 (produce_symtab): Update.
24914 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
24916 * config/s390/s390-opts.h (enum indirect_branch): Define.
24917 * config/s390/s390-protos.h (s390_return_addr_from_memory)
24918 (s390_indirect_branch_via_thunk)
24919 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
24920 (enum s390_indirect_branch_type): Define.
24921 * config/s390/s390.c (struct s390_frame_layout, struct
24922 machine_function): Remove.
24923 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
24924 (indirect_branch_table_label_no, indirect_branch_table_name):
24926 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
24927 (enum s390_indirect_branch_option): Define.
24928 (s390_return_addr_from_memory): New function.
24929 (s390_handle_string_attribute): New function.
24930 (s390_attribute_table): Add new attribute handler.
24931 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
24932 (s390_indirect_branch_via_thunk): New function.
24933 (s390_indirect_branch_via_inline_thunk): New function.
24934 (s390_function_ok_for_sibcall): When jumping via thunk disallow
24935 sibling call optimization for non z10 compiles.
24936 (s390_emit_call): Force indirect branch target to be a single
24937 register. Add r1 clobber for non-z10 compiles.
24938 (s390_emit_epilogue): Emit return jump via return_use expander.
24939 (s390_reorg): Handle JUMP_INSNs as execute targets.
24940 (s390_option_override_internal): Perform validity checks for the
24941 new command line options.
24942 (s390_indirect_branch_attrvalue): New function.
24943 (s390_indirect_branch_settings): New function.
24944 (s390_set_current_function): Invoke s390_indirect_branch_settings.
24945 (s390_output_indirect_thunk_function): New function.
24946 (s390_code_end): Implement target hook.
24947 (s390_case_values_threshold): Implement target hook.
24948 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
24950 * config/s390/s390.h (struct s390_frame_layout)
24951 (struct machine_function): Move here from s390.c.
24952 (TARGET_INDIRECT_BRANCH_NOBP_RET)
24953 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
24954 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
24955 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
24956 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
24957 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
24958 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
24959 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
24960 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
24961 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
24962 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
24963 (mnemonic attribute): Add values which aren't recognized
24965 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
24966 pattern for branch conversion. Fix mnemonic attribute.
24967 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
24968 indirect branch via thunk if requested.
24969 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
24970 ("*indirect_jump"): Disable for branch conversion using out of
24972 ("indirect_jump_via_thunk<mode>_z10")
24973 ("indirect_jump_via_thunk<mode>")
24974 ("indirect_jump_via_inlinethunk<mode>_z10")
24975 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
24976 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
24977 ("casesi_jump_via_inlinethunk<mode>_z10")
24978 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
24979 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
24980 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
24981 ("*indirect2_jump"): Disable for branch conversion.
24982 ("casesi_jump"): Turn into expander and expand patterns for branch
24984 ("return_use"): New expander.
24985 ("*return"): Emit return via thunk and rename it to ...
24986 ("*return<mode>"): ... this one.
24987 * config/s390/s390.opt: Add new options and and enum for the
24990 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
24992 * lra-constraints.c (match_reload): Unconditionally use
24993 gen_lowpart_SUBREG, rather than selecting between that
24994 and equivalent gen_rtx_SUBREG code.
24996 2018-02-08 Richard Biener <rguenther@suse.de>
24998 PR tree-optimization/84233
24999 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
25000 changed flag instead of boguously re-using phi_inserted.
25002 2018-02-08 Martin Jambor <mjambor@suse.cz>
25004 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
25005 static local variables.
25007 2018-02-08 Richard Biener <rguenther@suse.de>
25009 PR tree-optimization/84278
25010 * tree-vect-stmts.c (vectorizable_store): When looking for
25011 smaller vector types to perform grouped strided loads/stores
25012 make sure the mode is supported by the target.
25013 (vectorizable_load): Likewise.
25015 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
25017 * config/aarch64/aarch64.c (aarch64_components_for_bb):
25018 Increase LDP/STP opportunities by adding adjacent callee-saves.
25020 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
25022 PR rtl-optimization/84068
25023 PR rtl-optimization/83459
25024 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
25026 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
25028 PR tree-optimization/84224
25029 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
25030 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
25031 non-zero arguments.
25033 2018-02-07 Iain Sandoe <iain@codesourcery.com>
25036 * config/rs6000/altivec.md (*restore_world): Remove LR use.
25037 * config/rs6000/predicates.md (restore_world_operation): Adjust op
25038 count, remove one USE.
25040 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
25042 * doc/install.texi (Configuration): Document the
25043 --with-long-double-format={ibm,ieee} PowerPC configuration
25047 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
25048 Convert from define_expand to be define_insn_and_split. Rework
25049 float/double/_Float128 conversions to QI/HI/SImode to work with
25050 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
25051 conversions to QI/HImode types did a store and then a load to
25052 truncate the value. For conversions to VSX registers, don't split
25053 the insn, instead emit the code directly. Use the code iterator
25054 any_fix to combine signed and unsigned conversions.
25055 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
25056 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
25057 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
25058 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
25059 (fix_<mode>di2_hw): Likewise.
25060 (fixuns_<mode>di2_hw): Likewise.
25061 (fix_<mode>si2_hw): Likewise.
25062 (fixuns_<mode>si2_hw): Likewise.
25063 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
25064 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
25065 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
25066 fix<uns>_trunc<SFDF:mode>si2_p8.
25067 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
25068 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
25069 (fix<uns>_<mode>_mem): Likewise.
25070 (fctiw<u>z_<mode>_mem): Likewise.
25071 (fix<uns>_<mode>_mem): Likewise.
25072 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
25073 the register allocator from doing a direct move to the GPRs to do
25074 a store, and instead use the ISA 3.0 store byte/half-word from
25075 vector register instruction. For IEEE 128-bit floating point,
25076 also optimize stores of 32-bit ints.
25077 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
25079 2018-02-07 Alan Hayward <alan.hayward@arm.com>
25081 * genextract.c (push_pathstr_operand): New function to support
25083 (walk_rtx): Call push_pathstr_operand.
25084 (print_path): Support [a-zA-Z].
25086 2018-02-07 Richard Biener <rguenther@suse.de>
25088 PR tree-optimization/84037
25089 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
25090 (cse_and_gimplify_to_preheader): Declare.
25091 (vect_get_place_in_interleaving_chain): Likewise.
25092 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25094 (_loop_vec_info::~_loop_vec_info): Delete it.
25095 (cse_and_gimplify_to_preheader): New function.
25096 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
25097 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
25098 (vectorizable_load): Likewise. For grouped stores always base
25099 the IV on the first element.
25100 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
25101 condition before gimplifying.
25103 2018-02-07 Jakub Jelinek <jakub@redhat.com>
25105 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
25106 *DIV_EXPR and *MOD_EXPR.
25108 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
25111 * config/i386/i386.c (ix86_option_override_internal): Mask out
25112 the CF_SET bit when checking -fcf-protection.
25114 2018-02-07 Tom de Vries <tom@codesourcery.com>
25117 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
25120 2018-02-07 Richard Biener <rguenther@suse.de>
25122 PR tree-optimization/84204
25123 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
25126 PR tree-optimization/84205
25127 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
25128 special-case isl_ast_op_zdiv_r.
25130 PR tree-optimization/84223
25131 * graphite-scop-detection.c (gather_bbs::before_dom_children):
25132 Only add conditions from within the region.
25133 (gather_bbs::after_dom_children): Adjust.
25135 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
25138 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
25139 * config/avr/avr.md: Only post-reload split REG-REG moves if
25140 either register is GENERAL_REG_P.
25142 2018-02-07 Jakub Jelinek <jakub@redhat.com>
25144 PR tree-optimization/84235
25145 * tree-ssa-scopedtables.c
25146 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
25147 if the subtraction is performed in floating point type where NaNs are
25148 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
25149 build 1. Formatting fix.
25151 2018-02-06 Jakub Jelinek <jakub@redhat.com>
25154 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
25155 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
25156 and skip it regardless of bb boundaries. Use CALL_P macro,
25157 don't test INSN_P (insn) together with CALL_P or JUMP_P check
25158 unnecessarily, formatting fix.
25160 2018-02-06 Michael Collison <michael.collison@arm.com>
25162 * config/arm/thumb2.md:
25163 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
25164 (*thumb_mov_notscc): Ditto.
25166 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
25169 * config/rs6000/rs6000.md (su code attribute): Use "u" for
25170 unsigned_fix, not "s".
25172 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25174 * configure.ac (gcc_fn_eh_frame_ro): New function.
25175 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
25176 correct .eh_frame permissions.
25177 * configure: Regenerate.
25179 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
25181 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
25182 irrelevant options.
25184 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25186 * config/rs6000/rs6000.c (rs6000_option_override_internal):
25187 Display warning message for -mno-speculate-indirect-jumps.
25189 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
25191 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
25193 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
25195 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
25197 PR tree-optimization/84225
25198 * tree-eh.c (find_trapping_overflow): Only call
25199 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
25201 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
25204 * config/i386/i386.c: Reimplement the check of possible options
25205 -mibt/-mshstk conbination. Change error messages.
25206 * doc/invoke.texi: Fix a typo: remove extra '='.
25208 2018-02-06 Marek Polacek <polacek@redhat.com>
25210 PR tree-optimization/84228
25211 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
25213 2018-02-06 Tamar Christina <tamar.christina@arm.com>
25216 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
25217 emitted arch directives.
25218 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
25219 __ARM_FEATURE_COPROC before changing architectures.
25221 2018-02-06 Richard Biener <rguenther@suse.de>
25223 * config/i386/i386.c (print_reg): Fix typo.
25224 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
25226 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
25228 * configure: Regenerate.
25230 2018-02-05 Martin Sebor <msebor@redhat.com>
25232 PR tree-optimization/83369
25233 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
25236 2018-02-05 Martin Liska <mliska@suse.cz>
25238 * doc/invoke.texi: Cherry-pick upstream r323995.
25240 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
25242 * ira.c (ira_init_register_move_cost): Adjust comment.
25244 2018-02-05 Martin Liska <mliska@suse.cz>
25246 PR gcov-profile/84137
25247 * doc/gcov.texi: Fix typo in documentation.
25249 2018-02-05 Martin Liska <mliska@suse.cz>
25251 PR gcov-profile/83879
25252 * doc/gcov.texi: Document necessity of --dynamic-list-data when
25253 using dlopen functionality.
25255 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
25257 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
25258 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
25259 _mm_maskz_range_ss, _mm_mask_range_round_ss,
25260 _mm_maskz_range_round_ss): New intrinsics.
25261 (__builtin_ia32_rangesd128_round)
25262 (__builtin_ia32_rangess128_round): Remove.
25263 (__builtin_ia32_rangesd128_mask_round,
25264 __builtin_ia32_rangess128_mask_round): New builtins.
25265 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
25266 __builtin_ia32_rangess128_round): Remove.
25267 (__builtin_ia32_rangesd128_mask_round,
25268 __builtin_ia32_rangess128_mask_round): New builtins.
25269 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
25270 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
25271 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
25272 "<round_saeonly_constraint>")): Changed to ...
25273 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
25274 "<round_saeonly_scalar_constraint>")): ... this.
25275 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25276 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
25277 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
25278 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
25279 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
25281 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
25283 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
25285 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
25286 Remove all values except native, 8540 and 8548.
25288 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
25290 * config/i386/i386.c (ix86_output_function_return): Pass
25291 INVALID_REGNUM, instead of -1, as invalid register number to
25292 indirect_thunk_name and output_indirect_thunk.
25294 2018-02-02 Julia Koval <julia.koval@intel.com>
25296 * config.gcc: Add -march=icelake.
25297 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
25298 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
25299 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
25300 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
25301 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
25302 (processor_target_table): Add icelake.
25303 (ix86_option_override_internal): Handle new PTAs.
25304 (get_builtin_code_for_version): Handle icelake.
25305 (M_INTEL_COREI7_ICELAKE): New.
25306 (fold_builtin_cpu): Handle icelake.
25307 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
25308 * doc/invoke.texi: Add -march=icelake.
25310 2018-02-02 Julia Koval <julia.koval@intel.com>
25312 * config/i386/i386.c (ix86_option_override_internal): Change flags type
25313 to wide_int_bitmask.
25314 * wide-int-bitmask.h: New.
25316 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
25319 * config/i386/i386.md: Replace Pmode with word_mode in
25320 builtin_setjmp_setup and builtin_longjmp to support x32.
25322 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
25326 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
25328 (rs6000_supported_cpu_names): New static variable.
25329 (linux_cpu_translation_table): Likewise.
25330 (elf_platform) <cpu>: Define new static variable and use it.
25331 Translate kernel AT_PLATFORM name to canonical name if needed.
25332 Error if platform name is unknown.
25334 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
25337 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
25339 2018-02-01 Jeff Law <law@redhat.com>
25342 * config/i386/i386.c (release_scratch_register_on_entry): Add new
25343 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
25344 the scratch if RELEASE_VIA_POP is false.
25345 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
25346 If we have to save a temporary register, decrement SIZE appropriately.
25347 Pass new arguments to release_scratch_register_on_entry.
25348 (ix86_adjust_stack_and_probe): Likewise.
25349 (ix86_emit_probe_stack_range): Pass new arguments to
25350 release_scratch_register_on_entry.
25352 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
25354 PR rtl-optimization/84157
25355 * combine.c (change_zero_ext): Use REG_P predicate in
25356 front of HARD_REGISTER_P predicate.
25358 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
25360 * config/avr/avr.c (avr_option_override): Move disabling of
25361 -fdelete-null-pointer-checks to...
25362 * common/config/avr/avr-common.c (avr_option_optimization_table):
25365 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
25367 PR tree-optimization/81635
25368 * tree-data-ref.c (split_constant_offset_1): For types that
25369 wrap on overflow, try to use range info to prove that wrapping
25372 2018-02-01 Renlin Li <renlin.li@arm.com>
25375 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
25376 TAILCALL_ADDR_REGS.
25377 (aarch64_register_move_cost): Likewise.
25378 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
25379 TAILCALL_ADDR_REGS.
25380 (REG_CLASS_NAMES): Likewise.
25381 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
25382 TAILCALL_ADDR_REGS. Remove IP registers.
25383 * config/aarch64/aarch64.md (Ucs): Update register constraint.
25385 2018-02-01 Richard Biener <rguenther@suse.de>
25387 * domwalk.h (dom_walker::dom_walker): Add additional constructor
25388 for specifying RPO order and allow NULL for that.
25389 * domwalk.c (dom_walker::dom_walker): Likewise.
25390 (dom_walker::walk): Handle NULL RPO order.
25391 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
25393 (rewrite_update_dom_walker): Likewise.
25394 (mark_def_dom_walker): Likewise.
25396 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
25398 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
25399 (aarch64_maybe_expand_sve_subreg_move): Declare.
25400 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
25401 * config/aarch64/predicates.md (aarch64_any_register_operand): New
25403 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
25404 that are semantically a reverse operation.
25405 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
25406 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
25407 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
25409 (aarch64_can_change_mode_class): For big-endian, forbid changes
25410 between two SVE modes if they have different element sizes.
25412 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
25414 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
25415 the TImode handling for big-endian targets.
25417 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
25419 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
25420 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
25422 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
25423 Remove BSWAP handing for big-endian targets and use the form of
25424 LD1RQ appropariate for the mode.
25426 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
25428 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
25429 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
25430 duplicated element.
25432 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
25435 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
25436 check for operands that need to go through aarch64_sve_reload_be.
25438 2018-02-01 Jakub Jelinek <jakub@redhat.com>
25440 PR tree-optimization/81661
25441 PR tree-optimization/84117
25442 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
25443 * tree-eh.c: Include gimplify.h.
25444 (find_trapping_overflow, replace_trapping_overflow,
25445 rewrite_to_non_trapping_overflow): New functions.
25446 * tree-vect-loop.c: Include tree-eh.h.
25447 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
25448 * tree-data-ref.c: Include tree-eh.h.
25449 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
25451 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
25453 PR rtl-optimization/84123
25454 * combine.c (change_zero_ext): Check if hard register satisfies
25455 can_change_dest_mode before calling gen_lowpart_SUBREG.
25457 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
25460 * ira.c (ira_init_register_move_cost): Remove assert.
25462 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
25464 PR rtl-optimization/84071
25465 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
25466 * doc/tm.texi: Regenerate.
25468 2018-01-31 Richard Biener <rguenther@suse.de>
25470 PR tree-optimization/84132
25471 * tree-data-ref.c (analyze_miv_subscript): Properly
25472 check whether evolution_function_is_affine_multivariate_p
25473 before calling gcd_of_steps_may_divide_p.
25475 2018-01-31 Julia Koval <julia.koval@intel.com>
25478 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
25479 * config/i386/i386.md (rdpid_rex64) New.
25480 (rdpid): Make 32bit only.
25482 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
25485 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
25486 an IDENTIFIER_NODE for FUNCTION_TYPE's.
25488 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
25491 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
25493 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
25495 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
25497 PR rtl-optimization/84071
25498 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
25499 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
25501 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
25503 * config/arc/arc.c (arc_handle_aux_attribute): New function.
25504 (arc_attribute_table): Add 'aux' attribute.
25505 (arc_in_small_data_p): Consider aux like variables.
25506 (arc_is_aux_reg_p): New function.
25507 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
25508 (arc_get_aux_arg): New function.
25509 (prepare_move_operands): Handle aux-register access.
25510 (arc_handle_aux_attribute): New function.
25511 * doc/extend.texi (ARC Variable attributes): Add subsection.
25513 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
25515 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
25516 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
25517 (arc_attribute_table): Add 'uncached' attribute.
25518 (arc_print_operand): Print '.di' flag for uncached memory
25520 (arc_in_small_data_p): Do not consider for small data the uncached
25522 (arc_is_uncached_mem_p): New function.
25523 * config/arc/predicates.md (compact_store_memory_operand): Check
25524 for uncached memory accesses.
25525 (nonvol_nonimm_operand): Likewise.
25526 * doc/extend.texi (ARC Type Attribute): New subsection.
25528 2018-01-31 Jakub Jelinek <jakub@redhat.com>
25531 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
25532 falign-loops=): Add Optimization flag.
25534 2018-01-30 Jeff Law <law@redhat.com>
25537 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
25538 INT_REGISTERS_SAVED. Check it prior to calling
25539 get_scratch_register_on_entry.
25540 (ix86_adjust_stack_and_probe): Similarly.
25541 (ix86_emit_probe_stack_range): Similarly.
25542 (ix86_expand_prologue): Corresponding changes.
25544 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25547 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
25548 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
25550 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
25553 * lra-constraints.c (curr_insn_transform): Process AND in the
25556 2018-01-30 Jakub Jelinek <jakub@redhat.com>
25558 PR rtl-optimization/83986
25559 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
25560 dependence against last_pending_memory_flush in addition to
25561 pending_jump_insns.
25563 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
25565 PR tree-optimization/81611
25566 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
25569 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25572 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
25575 2018-01-30 Richard Biener <rguenther@suse.de>
25576 Jakub Jelinek <jakub@redhat.com>
25578 PR tree-optimization/84111
25579 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
25580 inner loops added during recursion, as they don't have up-to-date
25583 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
25586 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
25587 (can_inline_edge_by_limits_p): ... here.
25588 (can_early_inline_edge_p, check_callers,
25589 update_caller_keys, update_callee_keys, recursive_inlining,
25590 add_new_edges_to_heap, speculation_useful_p,
25591 inline_small_functions,
25592 inline_small_functions, flatten_function,
25593 inline_to_all_callers_1): Update.
25595 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
25597 * profile-count.c (profile_count::combine_with_ipa_count): Handle
25600 2018-01-30 Richard Biener <rguenther@suse.de>
25602 PR tree-optimization/83008
25603 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
25604 invariant and constant vector uses in stmts when they need
25605 more than one stmt.
25607 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25610 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
25611 * configure: Regenerate.
25613 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
25615 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
25617 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
25618 Use gen_rtx_REG rather than gen_lowpart.
25620 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
25622 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
25623 rather than 0 when creating partial subregs.
25625 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
25627 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
25630 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
25633 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
25634 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
25635 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
25636 flags. This restores the settings used before the 2017-07-24.
25637 Turning off pre increment/decrement/modify allows IVOPTS to
25638 optimize DF/SF loops where the index is an int.
25640 2018-01-29 Richard Biener <rguenther@suse.de>
25641 Kelvin Nilsen <kelvin@gcc.gnu.org>
25644 * tree-vect-stmts.c (vectorizable_call): Don't call
25645 targetm.vectorize_builtin_md_vectorized_function if callee is
25648 2018-01-22 Carl Love <cel@us.ibm.com>
25650 * doc/extend.tex: Fix typo in second arg in
25651 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
25653 2018-01-29 Richard Biener <rguenther@suse.de>
25655 PR tree-optimization/84086
25656 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
25657 (flush_ssaname_freelist): When SSA names were released reset
25658 the SCEV hash table.
25660 2018-01-29 Richard Biener <rguenther@suse.de>
25662 PR tree-optimization/84057
25663 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
25664 removed paths when removing edges.
25666 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
25668 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
25669 -mfunction-return=@var{choice}.
25671 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
25673 PR diagnostic/84034
25674 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
25675 Handle CR like TAB.
25676 (layout::print_source_line): Likewise.
25677 (test_get_line_width_without_trailing_whitespace): Add test cases.
25679 2018-01-27 Jakub Jelinek <jakub@redhat.com>
25681 PR middle-end/84040
25682 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
25685 2018-01-26 Jim Wilson <jimw@sifive.com>
25687 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
25689 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
25692 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25694 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
25695 and CMP + SUB-immediate -> SUBS.
25697 2018-01-26 Martin Sebor <msebor@redhat.com>
25699 PR tree-optimization/83896
25700 * tree-ssa-strlen.c (get_string_len): Rename...
25701 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
25702 Avoid assuming length is constant.
25703 (handle_char_store): Use HOST_WIDE_INT for string length.
25705 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
25708 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
25709 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
25711 2018-01-26 Richard Biener <rguenther@suse.de>
25713 PR rtl-optimization/84003
25714 * dse.c (record_store): Only record redundant stores when
25715 the earlier store aliases at least all accesses the later one does.
25717 2018-01-26 Jakub Jelinek <jakub@redhat.com>
25719 PR rtl-optimization/83985
25720 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
25721 REG_CFA_RESTORE insns.
25722 (delete_unmarked_insns): Don't ignore separate shrink wrapping
25723 REG_CFA_RESTORE insns here.
25726 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
25727 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
25729 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
25731 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
25732 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
25733 (arc_init): Likewise.
25734 (arc_override_options): Likewise.
25735 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
25737 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
25739 * config/arc/arc.h (TARGET_DBNZ): Define.
25740 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
25741 properly set the tune attribute.
25742 (dbnz): Use TARGET_DBNZ guard.
25743 * config/arc/arc.opt (mtune): Add core3 option.
25745 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
25747 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
25748 recognize new pic like addresses.
25749 (arc_delegitimize_address): Clean up.
25751 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
25753 * config/arc/arc-arches.def: Option mrf16 valid for all
25755 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
25756 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
25757 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
25758 * config/arc/arc-tables.opt: Regenerate.
25759 * config/arc/arc.c (arc_conditional_register_usage): Handle
25760 reduced register file case.
25761 (arc_file_start): Set must have build attributes.
25762 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
25763 mrf16 option value.
25764 * config/arc/arc.opt (mrf16): Add new option.
25765 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
25766 * config/arc/genmultilib.awk: Handle new mrf16 option.
25767 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
25768 * config/arc/t-multilib: Regenerate.
25769 * doc/invoke.texi (ARC Options): Document mrf16 option.
25771 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
25773 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
25774 * config/arc/arc.c (arc_handle_secure_attribute): New function.
25775 (arc_attribute_table): Add 'secure_call' attribute.
25776 (arc_print_operand): Print secure call operand.
25777 (arc_function_ok_for_sibcall): Don't optimize tail calls when
25779 (arc_is_secure_call_p): New function. * config/arc/arc.md
25780 (call_i): Add support for sjli instruction.
25781 (call_value_i): Likewise.
25782 * config/arc/constraints.md (Csc): New constraint.
25784 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
25785 John Eric Martin <John.Martin@emmicro-us.com>
25787 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
25788 * config/arc/arc.c (_arc_jli_section): New struct.
25789 (arc_jli_section): New type.
25790 (rc_jli_sections): New static variable.
25791 (arc_handle_jli_attribute): New function.
25792 (arc_attribute_table): Add jli_always and jli_fixed attribute.
25793 (arc_file_end): New function.
25794 (TARGET_ASM_FILE_END): Define.
25795 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
25796 (arc_add_jli_section): New function.
25797 (jli_call_scan): Likewise.
25798 (arc_reorg): Call jli_call_scan.
25799 (arc_output_addsi): Remove 'S' from printing asm operand.
25800 (arc_is_jli_call_p): New function.
25801 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
25803 (movhi_insn): Likewise.
25804 (movsi_insn): Likewise.
25805 (movsi_set_cc_insn): Likewise.
25806 (loadqi_update): Likewise.
25807 (load_zeroextendqisi_update): Likewise.
25808 (load_signextendqisi_update): Likewise.
25809 (loadhi_update): Likewise.
25810 (load_zeroextendhisi_update): Likewise.
25811 (load_signextendhisi_update): Likewise.
25812 (loadsi_update): Likewise.
25813 (loadsf_update): Likewise.
25814 (movsicc_insn): Likewise.
25815 (bset_insn): Likewise.
25816 (bxor_insn): Likewise.
25817 (bclr_insn): Likewise.
25818 (bmsk_insn): Likewise.
25819 (bicsi3_insn): Likewise.
25820 (cmpsi_cc_c_insn): Likewise.
25821 (movsi_ne): Likewise.
25822 (movsi_cond_exec): Likewise.
25823 (clrsbsi2): Likewise.
25824 (norm_f): Likewise.
25834 (call_i): Remove 'S' asm letter, add jli instruction.
25835 (call_value_i): Likewise.
25836 * config/arc/arc.op (mjli-always): New option.
25837 * config/arc/constraints.md (Cji): New constraint.
25838 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
25840 (subsf3_fpx): Likewise.
25841 (mulsf3_fpx): Likewise.
25842 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
25844 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
25845 function attrbutes.
25846 * doc/invoke.texi (ARC): Document mjli-always option.
25848 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
25850 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
25851 avoid addition with 0 and use incw and decw where possible.
25853 2018-01-26 Richard Biener <rguenther@suse.de>
25855 PR tree-optimization/81082
25856 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
25857 association if it requires casting to unsigned.
25858 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
25859 from fold_plusminus_mult_expr to catch important cases late when
25860 range info is available.
25862 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25864 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
25865 * configure.ac (hidden_linkonce): New test.
25866 * configure: Regenerate.
25867 * config.in: Regenerate.
25869 2018-01-26 Julia Koval <julia.koval@intel.com>
25871 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
25872 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
25873 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
25874 _mm_mask_bitshuffle_epi64_mask): Fix type.
25875 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
25876 USI_FTYPE_V4DI_V4DI_USI): Remove.
25877 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
25878 __builtin_ia32_vpshufbitqmb256_mask,
25879 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
25880 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
25881 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
25883 2018-01-26 Alan Modra <amodra@gmail.com>
25886 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
25887 UNSPEC_VBPERMQ. Sort other unspecs.
25889 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
25891 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
25893 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
25895 PR middle-end/83055
25896 * predict.c (drop_profile): Do not push/pop cfun; update also
25898 (handle_missing_profiles): Fix logic looking for zero profiles.
25900 2018-01-25 Jakub Jelinek <jakub@redhat.com>
25902 PR middle-end/83977
25903 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
25904 on functions with #pragma omp declare simd or functions with simd
25906 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
25907 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
25908 Remove trailing \n from warning_at calls.
25910 2018-01-25 Tom de Vries <tom@codesourcery.com>
25913 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
25914 for neutered workers.
25916 2018-01-24 Joseph Myers <joseph@codesourcery.com>
25919 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
25920 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
25922 2018-01-24 Jeff Law <law@redhat.com>
25925 * i386.c (get_probe_interval): Move to earlier point.
25926 (ix86_compute_frame_layout): If -fstack-clash-protection and
25927 the frame is larger than the probe interval, then use pushes
25928 to save registers rather than reg->mem moves.
25929 (ix86_expand_prologue): Remove conditional for int_registers_saved
25932 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
25935 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
25936 min/max for never referenced object.
25938 2018-01-24 Jakub Jelinek <jakub@redhat.com>
25940 PR middle-end/83977
25941 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
25943 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
25944 attributes from DECL_ATTRIBUTES (decl) without affecting
25945 DECL_ATTRIBUTES (current_function_decl).
25946 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
25947 functions with non-NULL DECL_ABSTRACT_ORIGIN.
25949 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
25951 PR tree-optimization/83979
25952 * fold-const.c (fold_comparison): Use constant_boolean_node
25953 instead of boolean_{true,false}_node.
25955 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
25957 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
25960 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25962 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
25963 Simplify the clause that sets the length attribute.
25964 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
25965 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
25966 clause that sets the length attribute.
25967 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
25969 2018-01-24 Tom de Vries <tom@codesourcery.com>
25972 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
25973 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
25974 Add strict parameter.
25975 (prevent_branch_around_nothing): Insert dummy insn between branch to
25976 label and label with no ptx insn inbetween.
25977 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
25979 2018-01-24 Tom de Vries <tom@codesourcery.com>
25982 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
25983 for neutered threads in warp.
25984 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
25986 2018-01-24 Richard Biener <rguenther@suse.de>
25988 PR tree-optimization/83176
25989 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
25992 2018-01-24 Richard Biener <rguenther@suse.de>
25994 PR tree-optimization/82819
25995 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
25996 code generating pluses that are no-ops in the target precision.
25998 2018-01-24 Richard Biener <rguenther@suse.de>
26000 PR middle-end/84000
26001 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
26003 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
26005 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
26006 to merge probabilities.
26007 * predict.c (probably_never_executed): Also mark as cold functions
26008 with global 0 profile and guessed local profile.
26009 * profile-count.c (profile_probability::combine_with_count): New
26011 * profile-count.h (profile_probability::operator*,
26012 profile_probability::operator*=, profile_probability::operator/,
26013 profile_probability::operator/=): Reduce precision to adjusted
26014 and set value to guessed on contradictory divisions.
26015 (profile_probability::combine_with_freq): Remove.
26016 (profile_probability::combine_wiht_count): Declare.
26017 (profile_count::force_nonzero):: Set to adjusted.
26018 (profile_count::probability_in):: Set quality to adjusted.
26019 * tree-ssa-tail-merge.c (replace_block_by): Use
26020 combine_with_count.
26022 2018-01-23 Andrew Waterman <andrew@sifive.com>
26023 Jim Wilson <jimw@sifive.com>
26025 * config/riscv/riscv.c (riscv_stack_boundary): New.
26026 (riscv_option_override): Set riscv_stack_boundary. Handle
26027 riscv_preferred_stack_boundary_arg.
26028 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
26029 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
26030 (STACK_BOUNDARY): Set to riscv_stack_boundary.
26031 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
26032 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
26033 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
26035 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
26038 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
26039 of struct ix86_frame.
26040 (ix86_expand_epilogue): Likewise. Add a local variable for
26041 the reg_save_offset field in struct ix86_frame.
26043 2018-01-23 Bin Cheng <bin.cheng@arm.com>
26045 PR tree-optimization/82604
26046 * tree-loop-distribution.c (enum partition_kind): New enum item
26047 PKIND_PARTIAL_MEMSET.
26048 (partition_builtin_p): Support above new enum item.
26049 (generate_code_for_partition): Ditto.
26050 (compute_access_range): Differentiate cases that equality can be
26051 proven at all loops, the innermost loops or no loops.
26052 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
26053 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
26054 (finalize_partitions, distribute_loop): Don't fuse partition of
26055 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
26056 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
26057 parloop is enabled.
26059 2018-01-23 Martin Liska <mliska@suse.cz>
26061 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
26062 order to ignore the predictor.
26063 (PRED_POLYMORPHIC_CALL): Likewise.
26064 (PRED_RECURSIVE_CALL): Likewise.
26066 2018-01-23 Martin Liska <mliska@suse.cz>
26068 * tree-profile.c (tree_profiling): Print function header to
26069 aware reader which function we are working on.
26070 * value-prof.c (gimple_find_values_to_profile): Do not print
26071 not interesting value histograms.
26073 2018-01-23 Martin Liska <mliska@suse.cz>
26075 * profile-count.h (enum profile_quality): Add
26076 profile_uninitialized as the first value. Do not number values
26077 as they are zero based.
26078 (profile_count::verify): Update sanity check.
26079 (profile_probability::verify): Likewise.
26081 2018-01-23 Nathan Sidwell <nathan@acm.org>
26083 * doc/invoke.texi (ffor-scope): Deprecate.
26085 2018-01-23 David Malcolm <dmalcolm@redhat.com>
26087 PR tree-optimization/83510
26088 * domwalk.c (set_all_edges_as_executable): New function.
26089 (dom_walker::dom_walker): Convert bool param
26090 "skip_unreachable_blocks" to enum reachability. Move setup of
26091 edge flags to set_all_edges_as_executable and only do it when
26092 reachability is REACHABLE_BLOCKS.
26093 * domwalk.h (enum dom_walker::reachability): New enum.
26094 (dom_walker::dom_walker): Convert bool param
26095 "skip_unreachable_blocks" to enum reachability.
26096 (set_all_edges_as_executable): New decl.
26097 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
26098 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
26100 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
26101 but converting true to REACHABLE_BLOCKS.
26102 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
26104 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
26105 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
26106 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
26108 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
26109 if check_all_array_refs will be called.
26111 2018-01-23 David Malcolm <dmalcolm@redhat.com>
26113 * tree.c (selftest::test_location_wrappers): Add more test
26116 2018-01-23 David Malcolm <dmalcolm@redhat.com>
26118 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
26119 (selftest::test_bit_in_range): Likewise.
26121 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
26124 * doc/sourcebuild.texi (vect_float): Say that the selector
26125 only describes the situation when -funsafe-math-optimizations is on.
26126 (vect_float_strict): Document.
26128 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
26130 PR tree-optimization/83965
26131 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
26132 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
26133 instead of checking only for a reduction.
26134 (vect_recog_widen_sum_pattern): Likewise.
26136 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
26138 * predict.c (probably_never_executed): Only use precise profile info.
26139 (compute_function_frequency): Skip after inlining hack since we now
26140 have quality checking.
26142 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
26144 * profile-count.h (profile_probability::very_unlikely,
26145 profile_probability::unlikely, profile_probability::even): Set
26146 precision to guessed.
26148 2018-01-23 Richard Biener <rguenther@suse.de>
26150 PR tree-optimization/83963
26151 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
26152 Properly terminate dominator walk when crossing the exit edge not
26153 when visiting its source block.
26155 2018-01-23 Jakub Jelinek <jakub@redhat.com>
26158 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
26159 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
26161 2018-01-22 Jakub Jelinek <jakub@redhat.com>
26163 PR tree-optimization/83957
26164 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
26165 semicolon after for body surrounded by braces.
26167 PR tree-optimization/83081
26168 * profile-count.h (profile_probability::split): New method.
26169 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
26170 Use profile_probability::split.
26171 (do_compare_rtx_and_jump): Fix adjustment of probabilities
26172 when splitting a single conditional jump into 2.
26174 2018-01-22 David Malcolm <dmalcolm@redhat.com>
26176 PR tree-optimization/69452
26177 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
26180 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
26182 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
26183 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
26184 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
26186 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
26188 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
26190 * config/rl78/rl78.md (movdi): New define_expand.
26191 * config/rl78/rl78.c (rl78_split_movdi): New function.
26193 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
26196 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
26198 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
26199 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
26200 128-bit to produce an UNSPEC move to get the double word with the
26201 signbit and then a shift directly to do signbit.
26202 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
26203 implementation with a new version that just does either a direct
26204 move or a regular move. Move memory interface to separate insns.
26205 Move insns so they are next to the expander.
26206 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
26207 with signbit move. Split big and little endian case.
26208 (signbit<mode>2_dm_mem_le): Likewise.
26209 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
26210 (signbit<mode>2_dm2): Likewise.
26212 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
26214 * config/rl78/rl78.md (anddi3): New define_expand.
26216 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
26218 * config/rl78/rl78.md (umindi3): New define_expand.
26220 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
26222 * config/rl78/rl78.md (smindi3): New define_expand.
26224 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
26226 * config/rl78/rl78.md (smaxdi3): New define_expand.
26228 2018-01-22 Carl Love <cel@us.ibm.com>
26230 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
26231 LVX_V1TI): Add macro expansion.
26232 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
26233 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
26234 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
26235 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
26236 Change check to determine if the instruction is a byte reversing
26237 entry. Fix typo in comment.
26238 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
26239 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
26240 Add def_builtin calls for new builtins.
26241 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
26242 Add define_insn expansion.
26244 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
26246 * config/rl78/rl78.md (umaxdi3): New define_expand.
26248 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
26250 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
26251 for non-QImode registers.
26253 2018-01-22 Richard Biener <rguenther@suse.de>
26255 PR tree-optimization/83963
26256 * graphite-scop-detection.c (scop_detection::get_sese): Delay
26257 including the loop exit block.
26258 (scop_detection::merge_sese): Likewise.
26259 (scop_detection::add_scop): Do it here instead.
26261 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26263 * doc/sourcebuild.texi (arm_softfloat): Document.
26265 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
26268 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
26269 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
26270 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
26272 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26273 David Edelsohn <dje.gcc@gmail.com>
26276 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
26277 Change "crset eq" to "crset 2".
26278 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
26279 (*call_indirect_aix<mode>_nospec): Likewise.
26280 (*call_value_indirect_aix<mode>_nospec): Likewise.
26281 (*call_indirect_elfv2<mode>_nospec): Likewise.
26282 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
26283 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
26284 change assembly output from . to $.
26285 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
26286 (indirect_jump<mode>_nospec): Change assembly output from . to $.
26287 (*tablejump<mode>_internal1_nospec): Likewise.
26289 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
26292 * config/sh/sh_optimize_sett_clrt.cc:
26293 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
26295 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
26297 PR tree-optimization/83940
26298 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
26299 offset_dt to vect_constant_def rather than vect_unknown_def_type.
26300 (vect_check_load_store_mask): Add a mask_dt_out parameter and
26301 use it to pass back the definition type.
26302 (vect_check_store_rhs): Likewise rhs_dt_out.
26303 (vect_build_gather_load_calls): Add a mask_dt argument and use
26304 it instead of a call to vect_is_simple_use.
26305 (vectorizable_store): Update calls to vect_check_load_store_mask
26306 and vect_check_store_rhs. Use the dt returned by the latter instead
26307 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
26308 instead of calls to vect_is_simple_use. Pass the scalar rather
26309 than the vector operand to vect_is_simple_use when handling
26310 second and subsequent copies of an rhs value.
26311 (vectorizable_load): Update calls to vect_check_load_store_mask
26312 and vect_build_gather_load_calls. Use the cached mask_dt and
26313 gs_info.offset_dt instead of calls to vect_is_simple_use.
26315 2018-01-20 Jakub Jelinek <jakub@redhat.com>
26317 PR middle-end/83945
26318 * tree-emutls.c: Include gimplify.h.
26319 (lower_emutls_2): New function.
26320 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
26321 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
26322 it before further processing.
26325 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
26326 UINTVAL (trueop1) instead of INTVAL (op1).
26328 2018-01-19 Jakub Jelinek <jakub@redhat.com>
26332 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
26333 INCOMING_FRAME_SP_OFFSET if not defined.
26334 (scan_trace): Add ENTRY argument. If true and
26335 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
26336 emit a note to adjust the CFA offset.
26337 (create_cfi_notes): Adjust scan_trace callers.
26338 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
26339 INCOMING_FRAME_SP_OFFSET in the CIE.
26340 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
26341 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
26343 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
26344 * doc/tm.texi: Regenerated.
26346 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
26348 PR rtl-optimization/83147
26349 * lra-constraints.c (remove_inheritance_pseudos): Use
26350 lra_substitute_pseudo_within_insn.
26352 2018-01-19 Tom de Vries <tom@codesourcery.com>
26353 Cesar Philippidis <cesar@codesourcery.com>
26356 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
26358 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
26361 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
26362 spaces for function labels.
26364 2018-01-19 Martin Liska <mliska@suse.cz>
26366 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
26367 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
26368 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
26369 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
26370 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
26371 (PRED_CONST_RETURN): Change from 69 to 65.
26372 (PRED_NULL_RETURN): Change from 91 to 71.
26373 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
26374 (PRED_LOOP_GUARD): Change from 66 to 73.
26376 2018-01-19 Martin Liska <mliska@suse.cz>
26378 * predict.c (predict_insn_def): Add new assert.
26379 (struct branch_predictor): Change type to signed integer.
26380 (test_prediction_value_range): Amend test to cover
26381 PROB_UNINITIALIZED.
26382 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
26383 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
26384 (PRED_LOOP_ITERATIONS_MAX): Likewise.
26385 (PRED_LOOP_IV_COMPARE): Likewise.
26386 * predict.h (PROB_UNINITIALIZED): Define new constant.
26388 2018-01-19 Martin Liska <mliska@suse.cz>
26390 * predict.c (dump_prediction): Add new format for
26391 analyze_brprob.py script which is enabled with -details
26393 * profile-count.h (precise_p): New function.
26395 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
26397 PR tree-optimization/83922
26398 * tree-vect-loop.c (vect_verify_full_masking): Return false if
26399 there are no statements that need masking.
26400 (vect_active_double_reduction_p): New function.
26401 (vect_analyze_loop_operations): Use it when handling phis that
26402 are not in the loop header.
26404 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
26406 PR tree-optimization/83914
26407 * tree-vect-loop.c (vectorizable_induction): Don't convert
26408 init_expr or apply the peeling adjustment for inductions
26409 that are nested within the vectorized loop.
26411 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26413 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
26416 2018-01-18 Jakub Jelinek <jakub@redhat.com>
26420 * function.h (gimplify_parameters): Add gimple_seq * argument.
26421 * function.c: Include gimple.h and options.h.
26422 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
26423 for the added local temporaries if needed.
26424 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
26425 if there are any parameter cleanups, wrap whole body into a
26426 try/finally with the cleanups.
26428 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
26431 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
26432 Use GET_MODE_CLASS for scalar floating point.
26434 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
26438 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
26439 Fix call of call_cgraph_insertion_hooks.
26441 2018-01-18 Martin Sebor <msebor@redhat.com>
26443 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
26445 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
26448 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
26451 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
26454 * common.opt: (-ffile-prefix-map): New option.
26455 * opts.c (common_handle_option): Defer it.
26456 * opts-global.c (handle_common_deferred_options): Handle it.
26457 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
26458 * file-prefix-map.h: New file.
26459 (remap_debug_filename, add_debug_prefix_map): ...here.
26460 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
26461 * final.c (debug_prefix_map, add_debug_prefix_map
26462 remap_debug_filename): Move to...
26463 * file-prefix-map.c: New file.
26464 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
26465 generalize, get rid of alloca(), use strrchr() instead of strchr().
26466 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
26467 Implement in terms of add_prefix_map().
26468 (remap_macro_filename, remap_debug_filename): Implement in term of
26470 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
26471 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
26472 * dbxout.c: Include file-prefix-map.h.
26473 * varasm.c: Likewise.
26474 * vmsdbgout.c: Likewise.
26475 * xcoffout.c: Likewise.
26476 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
26477 * doc/cppopts.texi (-fmacro-prefix-map): Document.
26478 * doc/invoke.texi (-ffile-prefix-map): Document.
26479 (-fdebug-prefix-map): Update description.
26481 2018-01-18 Martin Liska <mliska@suse.cz>
26483 * config/i386/i386.c (indirect_thunk_name): Document that also
26485 (output_indirect_thunk): Document why both instructions
26486 (pause and lfence) are generated.
26488 2018-01-18 Richard Biener <rguenther@suse.de>
26490 PR tree-optimization/83887
26491 * graphite-scop-detection.c
26492 (scop_detection::get_nearest_dom_with_single_entry): Remove.
26493 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
26494 (scop_detection::merge_sese): Re-implement with a flood-fill
26495 algorithm that properly finds a SESE region if it exists.
26497 2018-01-18 Jakub Jelinek <jakub@redhat.com>
26500 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
26501 pointer_diff optimizations use view_convert instead of convert.
26503 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26505 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
26506 Generate different code for -mno-speculate-indirect-jumps.
26507 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
26508 (*call_indirect_aix<mode>): Disable for
26509 -mno-speculate-indirect-jumps.
26510 (*call_indirect_aix<mode>_nospec): New define_insn.
26511 (*call_value_indirect_aix<mode>): Disable for
26512 -mno-speculate-indirect-jumps.
26513 (*call_value_indirect_aix<mode>_nospec): New define_insn.
26514 (*sibcall_nonlocal_sysv<mode>): Generate different code for
26515 -mno-speculate-indirect-jumps.
26516 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
26518 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
26520 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
26521 long double type, set the flags for noting the default long double
26522 type, even if we don't pass or return a long double type.
26524 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
26527 * ipa-inline.c (flatten_function): Do not overwrite final inlining
26530 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
26532 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
26533 support for merge[hl].
26534 (fold_mergehl_helper): New helper function.
26535 (tree-vector-builder.h): New #include for tree_vector_builder usage.
26536 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
26537 (altivec_vmrglw_direct): Add xxmrglw insn.
26539 2018-01-17 Andrew Waterman <andrew@sifive.com>
26541 * config/riscv/riscv.c (riscv_conditional_register_usage): If
26542 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
26544 2018-01-17 David Malcolm <dmalcolm@redhat.com>
26547 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
26548 call the lto_location_cache before reading the
26549 DECL_SOURCE_LOCATION of the types.
26551 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
26552 Richard Sandiford <richard.sandiford@linaro.org>
26554 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
26555 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
26556 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
26557 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
26558 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
26560 * config/aarch64/constraints.md (aarch64_movti_operand):
26562 * config/aarch64/predicates.md (Uti): Add new constraint.
26564 2018-01-17 Carl Love <cel@us.ibm.com>
26566 * config/rs6000/vsx.md (define_expand xl_len_r,
26567 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
26568 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
26570 (define_expand, define_insn): Move the shift left from the
26571 define_insn to the define_expand for lxvl and stxvl instructions.
26572 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
26573 and XL_LEN_R definitions to PURE.
26575 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
26577 * config/i386/i386.c (indirect_thunk_name): Declare regno
26578 as unsigned int. Compare regno with INVALID_REGNUM.
26579 (output_indirect_thunk): Ditto.
26580 (output_indirect_thunk_function): Ditto.
26581 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
26582 in the call to output_indirect_thunk_function.
26584 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
26586 PR middle-end/83884
26587 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
26588 rather than the size of inner_type to determine the stack slot size
26589 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
26591 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
26594 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
26597 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
26599 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
26600 endian Linux systems to optionally enable multilibs for selecting
26601 the long double type if the user configured an explicit type.
26602 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
26603 have no long double multilibs if not defined.
26604 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
26605 warn if the user used -mabi={ieee,ibm}longdouble and we built
26606 multilibs for long double.
26607 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
26608 appropriate multilib option.
26609 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
26611 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
26612 for building long double multilibs.
26613 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
26615 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
26617 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
26620 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
26622 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
26625 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
26628 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
26631 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
26633 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
26634 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
26636 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
26638 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
26639 different rtl trees depending on TARGET_64BIT.
26640 (rs6000_gen_lvx): Likewise.
26642 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
26644 * config/visium/visium.md (nop): Tweak comment.
26645 (hazard_nop): Likewise.
26647 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26649 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
26650 -mspeculate-indirect-jumps.
26651 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
26652 for -mno-speculate-indirect-jumps.
26653 (*call_indirect_elfv2<mode>_nospec): New define_insn.
26654 (*call_value_indirect_elfv2<mode>): Disable for
26655 -mno-speculate-indirect-jumps.
26656 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
26657 (indirect_jump): Emit different RTL for
26658 -mno-speculate-indirect-jumps.
26659 (*indirect_jump<mode>): Disable for
26660 -mno-speculate-indirect-jumps.
26661 (*indirect_jump<mode>_nospec): New define_insn.
26662 (tablejump): Emit different RTL for
26663 -mno-speculate-indirect-jumps.
26664 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
26665 (tablejumpsi_nospec): New define_expand.
26666 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
26667 (tablejumpdi_nospec): New define_expand.
26668 (*tablejump<mode>_internal1): Disable for
26669 -mno-speculate-indirect-jumps.
26670 (*tablejump<mode>_internal1_nospec): New define_insn.
26671 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
26674 2018-01-16 Artyom Skrobov tyomitch@gmail.com
26676 * caller-save.c (insert_save): Drop unnecessary parameter. All
26679 2018-01-16 Jakub Jelinek <jakub@redhat.com>
26680 Richard Biener <rguenth@suse.de>
26683 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
26684 return early, inline manually is_gimple_sizepos. Make sure if we
26685 call gimplify_expr we don't end up with a gimple constant.
26686 * tree.c (variably_modified_type_p): Don't return true for
26687 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
26688 * gimplify.h (is_gimple_sizepos): Remove.
26690 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
26692 PR tree-optimization/83857
26693 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
26694 vectorizable_live_operation for pure SLP statements.
26695 (vectorizable_live_operation): Handle PHIs.
26697 2018-01-16 Richard Biener <rguenther@suse.de>
26699 PR tree-optimization/83867
26700 * tree-vect-stmts.c (vect_transform_stmt): Precompute
26701 nested_in_vect_loop_p since the scalar stmt may get invalidated.
26703 2018-01-16 Jakub Jelinek <jakub@redhat.com>
26706 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
26707 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
26708 If off is not INTEGER_CST, issue a may not be aligned warning
26709 rather than isn't aligned. Use isn%'t rather than isn't.
26710 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
26712 <case MULT_EXPR>: Improve the case when bottom and one of the
26713 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
26714 operand, in that case check if the other operand is multiple of
26715 bottom divided by the INTEGER_CST operand.
26717 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
26720 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
26721 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
26722 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
26723 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
26724 * config/pa/pa.c (pa_function_arg_advance): Likewise.
26725 (pa_function_arg, pa_arg_partial_bytes): Likewise.
26726 (pa_function_arg_size): New function.
26728 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
26730 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
26731 in a separate statement.
26733 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
26735 PR tree-optimization/83847
26736 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
26737 group gathers and scatters.
26739 2018-01-16 Jakub Jelinek <jakub@redhat.com>
26741 PR rtl-optimization/86620
26742 * params.def (max-sched-ready-insns): Bump minimum value to 1.
26744 PR rtl-optimization/83213
26745 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
26746 to last if both are JUMP_INSNs.
26748 PR tree-optimization/83843
26749 * gimple-ssa-store-merging.c
26750 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
26751 store_immediate_info for bswap/nop orig_stores.
26753 2018-01-15 Andrew Waterman <andrew@sifive.com>
26755 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
26757 <UDIV>: Increase cost if !TARGET_DIV.
26759 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
26761 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
26762 (define_attr "cr_logical_3op"): New.
26763 (cceq_ior_compare): Adjust.
26764 (cceq_ior_compare_complement): Adjust.
26765 (*cceq_rev_compare): Adjust.
26766 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
26767 (is_cracked_insn): Adjust.
26768 (insn_must_be_first_in_group): Adjust.
26769 * config/rs6000/40x.md: Adjust.
26770 * config/rs6000/440.md: Adjust.
26771 * config/rs6000/476.md: Adjust.
26772 * config/rs6000/601.md: Adjust.
26773 * config/rs6000/603.md: Adjust.
26774 * config/rs6000/6xx.md: Adjust.
26775 * config/rs6000/7450.md: Adjust.
26776 * config/rs6000/7xx.md: Adjust.
26777 * config/rs6000/8540.md: Adjust.
26778 * config/rs6000/cell.md: Adjust.
26779 * config/rs6000/e300c2c3.md: Adjust.
26780 * config/rs6000/e500mc.md: Adjust.
26781 * config/rs6000/e500mc64.md: Adjust.
26782 * config/rs6000/e5500.md: Adjust.
26783 * config/rs6000/e6500.md: Adjust.
26784 * config/rs6000/mpc.md: Adjust.
26785 * config/rs6000/power4.md: Adjust.
26786 * config/rs6000/power5.md: Adjust.
26787 * config/rs6000/power6.md: Adjust.
26788 * config/rs6000/power7.md: Adjust.
26789 * config/rs6000/power8.md: Adjust.
26790 * config/rs6000/power9.md: Adjust.
26791 * config/rs6000/rs64.md: Adjust.
26792 * config/rs6000/titan.md: Adjust.
26794 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
26796 * config/i386/predicates.md (indirect_branch_operand): Rewrite
26797 ix86_indirect_branch_register logic.
26799 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
26801 * config/i386/constraints.md (Bs): Update
26802 ix86_indirect_branch_register check. Don't check
26803 ix86_indirect_branch_register with GOT_memory_operand.
26805 * config/i386/predicates.md (GOT_memory_operand): Don't check
26806 ix86_indirect_branch_register here.
26807 (GOT32_symbol_operand): Likewise.
26809 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
26811 * config/i386/predicates.md (constant_call_address_operand):
26812 Rewrite ix86_indirect_branch_register logic.
26813 (sibcall_insn_operand): Likewise.
26815 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
26817 * config/i386/constraints.md (Bs): Replace
26818 ix86_indirect_branch_thunk_register with
26819 ix86_indirect_branch_register.
26821 * config/i386/i386.md (indirect_jump): Likewise.
26822 (tablejump): Likewise.
26823 (*sibcall_memory): Likewise.
26824 (*sibcall_value_memory): Likewise.
26825 Peepholes of indirect call and jump via memory: Likewise.
26826 * config/i386/i386.opt: Likewise.
26827 * config/i386/predicates.md (indirect_branch_operand): Likewise.
26828 (GOT_memory_operand): Likewise.
26829 (call_insn_operand): Likewise.
26830 (sibcall_insn_operand): Likewise.
26831 (GOT32_symbol_operand): Likewise.
26833 2018-01-15 Jakub Jelinek <jakub@redhat.com>
26835 PR middle-end/83837
26836 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
26837 type rather than type addr's type points to.
26838 (expand_omp_atomic_mutex): Likewise.
26839 (expand_omp_atomic): Likewise.
26841 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
26844 * config/i386/i386.c (output_indirect_thunk_function): Use
26845 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
26846 for __x86_return_thunk.
26848 2018-01-15 Richard Biener <rguenther@suse.de>
26850 PR middle-end/83850
26851 * expmed.c (extract_bit_field_1): Fix typo.
26853 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26856 * config/arm/iterators.md (VF): New mode iterator.
26857 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
26858 Remove integer-related logic from pattern.
26859 (neon_vabd<mode>_3): Likewise.
26861 2018-01-15 Jakub Jelinek <jakub@redhat.com>
26863 PR middle-end/82694
26864 * common.opt (fstrict-overflow): No longer an alias.
26865 (fwrapv-pointer): New option.
26866 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
26867 also for pointer types based on flag_wrapv_pointer.
26868 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
26869 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
26870 opts->x_flag_wrapv got set.
26871 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
26872 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
26873 POINTER_TYPE_OVERFLOW_UNDEFINED.
26874 * match.pd: Likewise in address comparison pattern.
26875 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
26877 2018-01-15 Richard Biener <rguenther@suse.de>
26880 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
26881 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
26882 Reset type names to their identifier if their TYPE_DECL doesn't
26883 have linkage (and thus is used for ODR and devirt).
26884 (save_debug_info_for_decl): Remove.
26885 (save_debug_info_for_type): Likewise.
26886 (add_tree_to_fld_list): Adjust.
26887 * tree-pretty-print.c (dump_generic_node): Make dumping of
26888 type names more robust.
26890 2018-01-15 Richard Biener <rguenther@suse.de>
26892 * BASE-VER: Bump to 8.0.1.
26894 2018-01-14 Martin Sebor <msebor@redhat.com>
26897 * builtins.c (check_access): Avoid warning when the no-warning bit
26900 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
26902 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
26903 * ira-color (allocno_hard_regs_compare): Likewise.
26905 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
26908 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
26909 Use .pushsection/.popsection.
26911 2018-01-14 Martin Sebor <msebor@redhat.com>
26914 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
26916 2018-01-14 Jakub Jelinek <jakub@redhat.com>
26918 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
26919 entry from extra_headers.
26920 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
26921 extra_headers, make the list bitwise identical to the i?86-*-* one.
26923 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
26925 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
26926 -mcmodel=large with -mindirect-branch=thunk,
26927 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
26928 -mfunction-return=thunk-extern.
26929 * doc/invoke.texi: Document -mcmodel=large is incompatible with
26930 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
26931 -mfunction-return=thunk and -mfunction-return=thunk-extern.
26933 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
26935 * config/i386/i386.c (print_reg): Print the name of the full
26936 integer register without '%'.
26937 (ix86_print_operand): Handle 'V'.
26938 * doc/extend.texi: Document 'V' modifier.
26940 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
26942 * config/i386/constraints.md (Bs): Disallow memory operand for
26943 -mindirect-branch-register.
26945 * config/i386/predicates.md (indirect_branch_operand): Likewise.
26946 (GOT_memory_operand): Likewise.
26947 (call_insn_operand): Likewise.
26948 (sibcall_insn_operand): Likewise.
26949 (GOT32_symbol_operand): Likewise.
26950 * config/i386/i386.md (indirect_jump): Call convert_memory_address
26951 for -mindirect-branch-register.
26952 (tablejump): Likewise.
26953 (*sibcall_memory): Likewise.
26954 (*sibcall_value_memory): Likewise.
26955 Disallow peepholes of indirect call and jump via memory for
26956 -mindirect-branch-register.
26957 (*call_pop): Replace m with Bw.
26958 (*call_value_pop): Likewise.
26959 (*sibcall_pop_memory): Replace m with Bs.
26960 * config/i386/i386.opt (mindirect-branch-register): New option.
26961 * doc/invoke.texi: Document -mindirect-branch-register option.
26963 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
26965 * config/i386/i386-protos.h (ix86_output_function_return): New.
26966 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
26967 set function_return_type.
26968 (indirect_thunk_name): Add ret_p to indicate thunk for function
26970 (output_indirect_thunk_function): Pass false to
26971 indirect_thunk_name.
26972 (ix86_output_indirect_branch_via_reg): Likewise.
26973 (ix86_output_indirect_branch_via_push): Likewise.
26974 (output_indirect_thunk_function): Create alias for function
26975 return thunk if regno < 0.
26976 (ix86_output_function_return): New function.
26977 (ix86_handle_fndecl_attribute): Handle function_return.
26978 (ix86_attribute_table): Add function_return.
26979 * config/i386/i386.h (machine_function): Add
26980 function_return_type.
26981 * config/i386/i386.md (simple_return_internal): Use
26982 ix86_output_function_return.
26983 (simple_return_internal_long): Likewise.
26984 * config/i386/i386.opt (mfunction-return=): New option.
26985 (indirect_branch): Mention -mfunction-return=.
26986 * doc/extend.texi: Document function_return function attribute.
26987 * doc/invoke.texi: Document -mfunction-return= option.
26989 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
26991 * config/i386/i386-opts.h (indirect_branch): New.
26992 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
26993 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
26994 with local indirect jump when converting indirect call and jump.
26995 (ix86_set_indirect_branch_type): New.
26996 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
26997 (indirectlabelno): New.
26998 (indirect_thunk_needed): Likewise.
26999 (indirect_thunk_bnd_needed): Likewise.
27000 (indirect_thunks_used): Likewise.
27001 (indirect_thunks_bnd_used): Likewise.
27002 (INDIRECT_LABEL): Likewise.
27003 (indirect_thunk_name): Likewise.
27004 (output_indirect_thunk): Likewise.
27005 (output_indirect_thunk_function): Likewise.
27006 (ix86_output_indirect_branch_via_reg): Likewise.
27007 (ix86_output_indirect_branch_via_push): Likewise.
27008 (ix86_output_indirect_branch): Likewise.
27009 (ix86_output_indirect_jmp): Likewise.
27010 (ix86_code_end): Call output_indirect_thunk_function if needed.
27011 (ix86_output_call_insn): Call ix86_output_indirect_branch if
27013 (ix86_handle_fndecl_attribute): Handle indirect_branch.
27014 (ix86_attribute_table): Add indirect_branch.
27015 * config/i386/i386.h (machine_function): Add indirect_branch_type
27016 and has_local_indirect_jump.
27017 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
27019 (tablejump): Likewise.
27020 (*indirect_jump): Use ix86_output_indirect_jmp.
27021 (*tablejump_1): Likewise.
27022 (simple_return_indirect_internal): Likewise.
27023 * config/i386/i386.opt (mindirect-branch=): New option.
27024 (indirect_branch): New.
27027 (thunk-inline): Likewise.
27028 (thunk-extern): Likewise.
27029 * doc/extend.texi: Document indirect_branch function attribute.
27030 * doc/invoke.texi: Document -mindirect-branch= option.
27032 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
27035 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
27037 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
27039 * ipa-inline.c (want_inline_small_function_p): Return false if
27040 inlining has already failed with CIF_FINAL_ERROR.
27041 (update_caller_keys): Call want_inline_small_function_p before
27043 (update_callee_keys): Likewise.
27045 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
27047 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
27049 (rs6000_quadword_masked_address_p): Likewise.
27050 (quad_aligned_load_p): Likewise.
27051 (quad_aligned_store_p): Likewise.
27052 (const_load_sequence_p): Add comment to describe the outer-most loop.
27053 (mimic_memory_attributes_and_flags): New function.
27054 (rs6000_gen_stvx): Likewise.
27055 (replace_swapped_aligned_store): Likewise.
27056 (rs6000_gen_lvx): Likewise.
27057 (replace_swapped_aligned_load): Likewise.
27058 (replace_swapped_load_constant): Capitalize argument name in
27059 comment describing this function.
27060 (rs6000_analyze_swaps): Add a third pass to search for vector loads
27061 and stores that access quad-word aligned addresses and replace
27062 with stvx or lvx instructions when appropriate.
27063 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
27064 New function prototype.
27065 (rs6000_quadword_masked_address_p): Likewise.
27066 (rs6000_gen_lvx): Likewise.
27067 (rs6000_gen_stvx): Likewise.
27068 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
27069 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
27070 when memory address is aligned.
27071 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
27072 this split to select lvx instruction when memory address is aligned.
27073 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
27074 instruction when memory address is aligned.
27075 (*vsx_le_perm_load_v16qi): Likewise.
27076 (four unnamed splitters): Modify to select the stvx instruction
27077 when memory is aligned.
27079 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
27081 * predict.c (determine_unlikely_bbs): Handle correctly BBs
27082 which appears in the queue multiple times.
27084 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27085 Alan Hayward <alan.hayward@arm.com>
27086 David Sherwood <david.sherwood@arm.com>
27088 * tree-vectorizer.h (vec_lower_bound): New structure.
27089 (_loop_vec_info): Add check_nonzero and lower_bounds.
27090 (LOOP_VINFO_CHECK_NONZERO): New macro.
27091 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
27092 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
27093 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
27094 fields. Make seg_len the distance travelled, not including the
27096 (dr_direction_indicator): Declare.
27097 (dr_zero_step_indicator): Likewise.
27098 (dr_known_forward_stride_p): Likewise.
27099 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
27101 (runtime_alias_check_p): Allow runtime alias checks with
27103 (operator ==): Compare access_size and align.
27104 (prune_runtime_alias_test_list): Rework for new distinction between
27105 the access_size and seg_len.
27106 (create_intersect_range_checks_index): Likewise. Cope with polynomial
27108 (get_segment_min_max): New function.
27109 (create_intersect_range_checks): Use it.
27110 (dr_step_indicator): New function.
27111 (dr_direction_indicator): Likewise.
27112 (dr_zero_step_indicator): Likewise.
27113 (dr_known_forward_stride_p): Likewise.
27114 * tree-loop-distribution.c (data_ref_segment_size): Return
27115 DR_STEP * (niters - 1).
27116 (compute_alias_check_pairs): Update call to the dr_with_seg_len
27118 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
27119 (vect_preserves_scalar_order_p): New function, split out from...
27120 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
27121 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
27122 (vect_vfa_access_size): New function.
27123 (vect_vfa_align): Likewise.
27124 (vect_compile_time_alias): Take access_size_a and access_b arguments.
27125 (dump_lower_bound): New function.
27126 (vect_check_lower_bound): Likewise.
27127 (vect_small_gap_p): Likewise.
27128 (vectorizable_with_step_bound_p): Likewise.
27129 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
27130 depencies if the vectorization factor is 1. Convert the checks
27131 for nonzero steps into checks on the bounds of DR_STEP. Try using
27132 a bunds check for variable steps if the minimum required step is
27133 relatively small. Update calls to the dr_with_seg_len
27134 constructor and to vect_compile_time_alias.
27135 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
27137 (vect_loop_versioning): Call it.
27138 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
27140 (vect_estimate_min_profitable_iters): Account for any bounds checks.
27142 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27143 Alan Hayward <alan.hayward@arm.com>
27144 David Sherwood <david.sherwood@arm.com>
27146 * doc/sourcebuild.texi (vect_scatter_store): Document.
27147 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
27149 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
27151 * genopinit.c (main): Add supports_vec_scatter_store and
27152 supports_vec_scatter_store_cached to target_optabs.
27153 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
27154 IFN_MASK_SCATTER_STORE.
27155 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
27157 * internal-fn.h (internal_store_fn_p): Declare.
27158 (internal_fn_stored_value_index): Likewise.
27159 * internal-fn.c (scatter_store_direct): New macro.
27160 (expand_scatter_store_optab_fn): New function.
27161 (direct_scatter_store_optab_supported_p): New macro.
27162 (internal_store_fn_p): New function.
27163 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
27164 IFN_MASK_SCATTER_STORE.
27165 (internal_fn_mask_index): Likewise.
27166 (internal_fn_stored_value_index): New function.
27167 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
27168 for scatter stores.
27169 * optabs-query.h (supports_vec_scatter_store_p): Declare.
27170 * optabs-query.c (supports_vec_scatter_store_p): New function.
27171 * tree-vectorizer.h (vect_get_store_rhs): Declare.
27172 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
27173 true for scatter stores.
27174 (vect_gather_scatter_fn_p): Handle scatter stores too.
27175 (vect_check_gather_scatter): Consider using scatter stores if
27176 supports_vec_scatter_store_p.
27177 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
27178 scatter stores too.
27179 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
27180 internal_fn_stored_value_index.
27181 (check_load_store_masking): Handle scatter stores too.
27182 (vect_get_store_rhs): Make public.
27183 (vectorizable_call): Use internal_store_fn_p.
27184 (vectorizable_store): Handle scatter store internal functions.
27185 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
27186 when deciding whether the end of the group has been reached.
27187 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
27188 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
27189 (mask_scatter_store<mode>): New insns.
27191 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27192 Alan Hayward <alan.hayward@arm.com>
27193 David Sherwood <david.sherwood@arm.com>
27195 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
27196 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
27197 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
27199 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
27200 Use vect_truncate_gather_scatter_offset if we can't treat the
27201 operation as a normal gather load or scatter store.
27202 (get_group_load_store_type): Take the gather_scatter_info
27203 as argument. Try using a gather load or scatter store for
27204 single-element groups.
27205 (get_load_store_type): Update calls to get_group_load_store_type
27206 and vect_use_strided_gather_scatters_p.
27208 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27209 Alan Hayward <alan.hayward@arm.com>
27210 David Sherwood <david.sherwood@arm.com>
27212 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
27213 optional tree argument.
27214 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
27216 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
27217 but continue to use the current value as a fallback.
27218 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
27219 to compare the updates.
27220 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
27221 (get_load_store_type): Use it when handling a strided access.
27222 (vect_get_strided_load_store_ops): New function.
27223 (vect_get_data_ptr_increment): Likewise.
27224 (vectorizable_load): Handle strided gather loads. Always pass
27225 a step to vect_create_data_ref_ptr and bump_vector_ptr.
27227 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27228 Alan Hayward <alan.hayward@arm.com>
27229 David Sherwood <david.sherwood@arm.com>
27231 * doc/md.texi (gather_load@var{m}): Document.
27232 (mask_gather_load@var{m}): Likewise.
27233 * genopinit.c (main): Add supports_vec_gather_load and
27234 supports_vec_gather_load_cached to target_optabs.
27235 * optabs-tree.c (init_tree_optimization_optabs): Use
27236 ggc_cleared_alloc to allocate target_optabs.
27237 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
27238 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
27240 * internal-fn.h (internal_load_fn_p): Declare.
27241 (internal_gather_scatter_fn_p): Likewise.
27242 (internal_fn_mask_index): Likewise.
27243 (internal_gather_scatter_fn_supported_p): Likewise.
27244 * internal-fn.c (gather_load_direct): New macro.
27245 (expand_gather_load_optab_fn): New function.
27246 (direct_gather_load_optab_supported_p): New macro.
27247 (direct_internal_fn_optab): New function.
27248 (internal_load_fn_p): Likewise.
27249 (internal_gather_scatter_fn_p): Likewise.
27250 (internal_fn_mask_index): Likewise.
27251 (internal_gather_scatter_fn_supported_p): Likewise.
27252 * optabs-query.c (supports_at_least_one_mode_p): New function.
27253 (supports_vec_gather_load_p): Likewise.
27254 * optabs-query.h (supports_vec_gather_load_p): Declare.
27255 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
27256 and memory_type field.
27257 (NUM_PATTERNS): Bump to 15.
27258 * tree-vect-data-refs.c: Include internal-fn.h.
27259 (vect_gather_scatter_fn_p): New function.
27260 (vect_describe_gather_scatter_call): Likewise.
27261 (vect_check_gather_scatter): Try using internal functions for
27262 gather loads. Recognize existing calls to a gather load function.
27263 (vect_analyze_data_refs): Consider using gather loads if
27264 supports_vec_gather_load_p.
27265 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
27266 (vect_get_gather_scatter_offset_type): Likewise.
27267 (vect_convert_mask_for_vectype): Likewise.
27268 (vect_add_conversion_to_patterm): Likewise.
27269 (vect_try_gather_scatter_pattern): Likewise.
27270 (vect_recog_gather_scatter_pattern): New pattern recognizer.
27271 (vect_vect_recog_func_ptrs): Add it.
27272 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
27273 internal_fn_mask_index and internal_gather_scatter_fn_p.
27274 (check_load_store_masking): Take the gather_scatter_info as an
27275 argument and handle gather loads.
27276 (vect_get_gather_scatter_ops): New function.
27277 (vectorizable_call): Check internal_load_fn_p.
27278 (vectorizable_load): Likewise. Handle gather load internal
27280 (vectorizable_store): Update call to check_load_store_masking.
27281 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
27282 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
27283 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
27284 (aarch64_gather_scale_operand_d): New predicates.
27285 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
27286 (mask_gather_load<mode>): New insns.
27288 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27289 Alan Hayward <alan.hayward@arm.com>
27290 David Sherwood <david.sherwood@arm.com>
27292 * optabs.def (fold_left_plus_optab): New optab.
27293 * doc/md.texi (fold_left_plus_@var{m}): Document.
27294 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
27295 * internal-fn.c (fold_left_direct): Define.
27296 (expand_fold_left_optab_fn): Likewise.
27297 (direct_fold_left_optab_supported_p): Likewise.
27298 * fold-const-call.c (fold_const_fold_left): New function.
27299 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
27300 * tree-parloops.c (valid_reduction_p): New function.
27301 (gather_scalar_reductions): Use it.
27302 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
27303 (vect_finish_replace_stmt): Declare.
27304 * tree-vect-loop.c (fold_left_reduction_fn): New function.
27305 (needs_fold_left_reduction_p): New function, split out from...
27306 (vect_is_simple_reduction): ...here. Accept reductions that
27307 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
27308 (vect_force_simple_reduction): Also store the reduction type in
27309 the assignment's STMT_VINFO_REDUC_TYPE.
27310 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
27311 (merge_with_identity): New function.
27312 (vect_expand_fold_left): Likewise.
27313 (vectorize_fold_left_reduction): Likewise.
27314 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
27315 scalar phi in place for it. Check for target support and reject
27316 cases that would reassociate the operation. Defer the transform
27317 phase to vectorize_fold_left_reduction.
27318 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
27319 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
27320 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
27322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27324 * tree-if-conv.c (predicate_mem_writes): Remove redundant
27325 call to ifc_temp_var.
27327 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27328 Alan Hayward <alan.hayward@arm.com>
27329 David Sherwood <david.sherwood@arm.com>
27331 * target.def (legitimize_address_displacement): Take the original
27332 offset as a poly_int.
27333 * targhooks.h (default_legitimize_address_displacement): Update
27335 * targhooks.c (default_legitimize_address_displacement): Likewise.
27336 * doc/tm.texi: Regenerate.
27337 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
27338 as an argument, moving assert of ad->disp == ad->disp_term to...
27339 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
27340 Try calling targetm.legitimize_address_displacement before expanding
27341 the address rather than afterwards, and adjust for the new interface.
27342 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
27343 Match the new hook interface. Handle SVE addresses.
27344 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
27345 new hook interface.
27347 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27349 * Makefile.in (OBJS): Add early-remat.o.
27350 * target.def (select_early_remat_modes): New hook.
27351 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
27352 * doc/tm.texi: Regenerate.
27353 * targhooks.h (default_select_early_remat_modes): Declare.
27354 * targhooks.c (default_select_early_remat_modes): New function.
27355 * timevar.def (TV_EARLY_REMAT): New timevar.
27356 * passes.def (pass_early_remat): New pass.
27357 * tree-pass.h (make_pass_early_remat): Declare.
27358 * early-remat.c: New file.
27359 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
27361 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
27363 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27364 Alan Hayward <alan.hayward@arm.com>
27365 David Sherwood <david.sherwood@arm.com>
27367 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
27368 vfm1 with a bound_epilog parameter.
27369 (vect_do_peeling): Update calls accordingly, and move the prologue
27370 call earlier in the function. Treat the base bound_epilog as 0 for
27371 fully-masked loops and retain vf - 1 for other loops. Add 1 to
27372 this base when peeling for gaps.
27373 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
27374 with fully-masked loops.
27375 (vect_estimate_min_profitable_iters): Handle the single peeled
27376 iteration in that case.
27378 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27379 Alan Hayward <alan.hayward@arm.com>
27380 David Sherwood <david.sherwood@arm.com>
27382 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
27383 single-element interleaving even if the size is not a power of 2.
27384 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
27385 accesses for single-element interleaving if the group size is
27388 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27389 Alan Hayward <alan.hayward@arm.com>
27390 David Sherwood <david.sherwood@arm.com>
27392 * doc/md.texi (fold_extract_last_@var{m}): Document.
27393 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
27394 * optabs.def (fold_extract_last_optab): New optab.
27395 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
27396 * internal-fn.c (fold_extract_direct): New macro.
27397 (expand_fold_extract_optab_fn): Likewise.
27398 (direct_fold_extract_optab_supported_p): Likewise.
27399 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
27400 * tree-vect-loop.c (vect_model_reduction_cost): Handle
27401 EXTRACT_LAST_REDUCTION.
27402 (get_initial_def_for_reduction): Do not create an initial vector
27403 for EXTRACT_LAST_REDUCTION reductions.
27404 (vectorizable_reduction): Leave the scalar phi in place for
27405 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
27406 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
27407 epilogue code for EXTRACT_LAST_REDUCTION and defer the
27408 transform phase to vectorizable_condition.
27409 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
27411 (vect_finish_stmt_generation): ...here.
27412 (vect_finish_replace_stmt): New function.
27413 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
27414 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
27416 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
27418 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27419 Alan Hayward <alan.hayward@arm.com>
27420 David Sherwood <david.sherwood@arm.com>
27422 * doc/md.texi (extract_last_@var{m}): Document.
27423 * optabs.def (extract_last_optab): New optab.
27424 * internal-fn.def (EXTRACT_LAST): New internal function.
27425 * internal-fn.c (cond_unary_direct): New macro.
27426 (expand_cond_unary_optab_fn): Likewise.
27427 (direct_cond_unary_optab_supported_p): Likewise.
27428 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
27429 loops using EXTRACT_LAST.
27430 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
27431 (extract_last_<mode>): ...this optab.
27432 (vec_extract<mode><Vel>): Update accordingly.
27434 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27435 Alan Hayward <alan.hayward@arm.com>
27436 David Sherwood <david.sherwood@arm.com>
27438 * target.def (empty_mask_is_expensive): New hook.
27439 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
27440 * doc/tm.texi: Regenerate.
27441 * targhooks.h (default_empty_mask_is_expensive): Declare.
27442 * targhooks.c (default_empty_mask_is_expensive): New function.
27443 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
27444 if the target says that empty masks are expensive.
27445 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
27447 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
27449 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27450 Alan Hayward <alan.hayward@arm.com>
27451 David Sherwood <david.sherwood@arm.com>
27453 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
27454 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
27455 (vect_use_loop_mask_for_alignment_p): New function.
27456 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
27457 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
27458 niters_skip argument. Make sure that the first niters_skip elements
27459 of the first iteration are inactive.
27460 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
27461 Update call to vect_set_loop_masks_directly.
27462 (get_misalign_in_elems): New function, split out from...
27463 (vect_gen_prolog_loop_niters): ...here.
27464 (vect_update_init_of_dr): Take a code argument that specifies whether
27465 the adjustment should be added or subtracted.
27466 (vect_update_init_of_drs): Likewise.
27467 (vect_prepare_for_masked_peels): New function.
27468 (vect_do_peeling): Skip prologue peeling if we're using a mask
27469 instead. Update call to vect_update_inits_of_drs.
27470 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
27472 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
27473 alignment. Do not include the number of peeled iterations in
27474 the minimum threshold in that case.
27475 (vectorizable_induction): Adjust the start value down by
27476 LOOP_VINFO_MASK_SKIP_NITERS iterations.
27477 (vect_transform_loop): Call vect_prepare_for_masked_peels.
27478 Take the number of skipped iterations into account when calculating
27480 * tree-vect-stmts.c (vect_gen_while_not): New function.
27482 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27483 Alan Hayward <alan.hayward@arm.com>
27484 David Sherwood <david.sherwood@arm.com>
27486 * doc/sourcebuild.texi (vect_fully_masked): Document.
27487 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
27488 default value to 0.
27489 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
27491 (vect_analyze_loop_2): ...here. Don't check the vectorization
27492 factor against the number of loop iterations if the loop is
27495 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27496 Alan Hayward <alan.hayward@arm.com>
27497 David Sherwood <david.sherwood@arm.com>
27499 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
27500 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
27501 (dump_groups): Update accordingly.
27502 (iv_use::mem_type): New member variable.
27503 (address_p): New function.
27504 (record_use): Add a mem_type argument and initialize the new
27506 (record_group_use): Add a mem_type argument. Use address_p.
27507 Remove obsolete null checks of base_object. Update call to record_use.
27508 (find_interesting_uses_op): Update call to record_group_use.
27509 (find_interesting_uses_cond): Likewise.
27510 (find_interesting_uses_address): Likewise.
27511 (get_mem_type_for_internal_fn): New function.
27512 (find_address_like_use): Likewise.
27513 (find_interesting_uses_stmt): Try find_address_like_use before
27514 calling find_interesting_uses_op.
27515 (addr_offset_valid_p): Use the iv mem_type field as the type
27516 of the addressed memory.
27517 (add_autoinc_candidates): Likewise.
27518 (get_address_cost): Likewise.
27519 (split_small_address_groups_p): Use address_p.
27520 (split_address_groups): Likewise.
27521 (add_iv_candidate_for_use): Likewise.
27522 (autoinc_possible_for_pair): Likewise.
27523 (rewrite_groups): Likewise.
27524 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
27525 (determine_group_iv_cost): Update after split of USE_ADDRESS.
27526 (get_alias_ptr_type_for_ptr_address): New function.
27527 (rewrite_use_address): Rewrite address uses in calls that were
27528 identified by find_address_like_use.
27530 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27531 Alan Hayward <alan.hayward@arm.com>
27532 David Sherwood <david.sherwood@arm.com>
27534 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
27536 * gimple-expr.h (is_gimple_addressable: Likewise.
27537 * gimple-expr.c (is_gimple_address): Likewise.
27538 * internal-fn.c (expand_call_mem_ref): New function.
27539 (expand_mask_load_optab_fn): Use it.
27540 (expand_mask_store_optab_fn): Likewise.
27542 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27543 Alan Hayward <alan.hayward@arm.com>
27544 David Sherwood <david.sherwood@arm.com>
27546 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
27547 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
27548 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
27549 (cond_umax@var{mode}): Document.
27550 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
27551 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
27552 (cond_umin_optab, cond_umax_optab): New optabs.
27553 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
27554 (COND_IOR, COND_XOR): New internal functions.
27555 * internal-fn.h (get_conditional_internal_fn): Declare.
27556 * internal-fn.c (cond_binary_direct): New macro.
27557 (expand_cond_binary_optab_fn): Likewise.
27558 (direct_cond_binary_optab_supported_p): Likewise.
27559 (get_conditional_internal_fn): New function.
27560 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
27561 Cope with reduction statements that are vectorized as calls rather
27563 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
27564 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
27565 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
27566 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
27567 (UNSPEC_COND_EOR): New unspecs.
27568 (optab): Add mappings for them.
27569 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
27570 (sve_int_op, sve_fp_op): New int attributes.
27572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27573 Alan Hayward <alan.hayward@arm.com>
27574 David Sherwood <david.sherwood@arm.com>
27576 * optabs.def (while_ult_optab): New optab.
27577 * doc/md.texi (while_ult@var{m}@var{n}): Document.
27578 * internal-fn.def (WHILE_ULT): New internal function.
27579 * internal-fn.h (direct_internal_fn_supported_p): New override
27580 that takes two types as argument.
27581 * internal-fn.c (while_direct): New macro.
27582 (expand_while_optab_fn): New function.
27583 (convert_optab_supported_p): Likewise.
27584 (direct_while_optab_supported_p): New macro.
27585 * wide-int.h (wi::udiv_ceil): New function.
27586 * tree-vectorizer.h (rgroup_masks): New structure.
27587 (vec_loop_masks): New typedef.
27588 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
27589 and fully_masked_p.
27590 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
27591 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
27592 (vect_max_vf): New function.
27593 (slpeel_make_loop_iterate_ntimes): Delete.
27594 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
27595 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
27596 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
27597 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
27598 internal-fn.h, stor-layout.h and optabs-query.h.
27599 (vect_set_loop_mask): New function.
27600 (add_preheader_seq): Likewise.
27601 (add_header_seq): Likewise.
27602 (interleave_supported_p): Likewise.
27603 (vect_maybe_permute_loop_masks): Likewise.
27604 (vect_set_loop_masks_directly): Likewise.
27605 (vect_set_loop_condition_masked): Likewise.
27606 (vect_set_loop_condition_unmasked): New function, split out from
27607 slpeel_make_loop_iterate_ntimes.
27608 (slpeel_make_loop_iterate_ntimes): Rename to..
27609 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
27610 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
27611 (vect_do_peeling): Update call accordingly.
27612 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
27614 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
27615 mask_compare_type, can_fully_mask_p and fully_masked_p.
27616 (release_vec_loop_masks): New function.
27617 (_loop_vec_info): Use it to free the loop masks.
27618 (can_produce_all_loop_masks_p): New function.
27619 (vect_get_max_nscalars_per_iter): Likewise.
27620 (vect_verify_full_masking): Likewise.
27621 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
27622 retries, and free the mask rgroups before retrying. Check loop-wide
27623 reasons for disallowing fully-masked loops. Make the final decision
27624 about whether use a fully-masked loop or not.
27625 (vect_estimate_min_profitable_iters): Do not assume that peeling
27626 for the number of iterations will be needed for fully-masked loops.
27627 (vectorizable_reduction): Disable fully-masked loops.
27628 (vectorizable_live_operation): Likewise.
27629 (vect_halve_mask_nunits): New function.
27630 (vect_double_mask_nunits): Likewise.
27631 (vect_record_loop_mask): Likewise.
27632 (vect_get_loop_mask): Likewise.
27633 (vect_transform_loop): Handle the case in which the final loop
27634 iteration might handle a partial vector. Call vect_set_loop_condition
27635 instead of slpeel_make_loop_iterate_ntimes.
27636 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
27637 (check_load_store_masking): New function.
27638 (prepare_load_store_mask): Likewise.
27639 (vectorizable_store): Handle fully-masked loops.
27640 (vectorizable_load): Likewise.
27641 (supportable_widening_operation): Use vect_halve_mask_nunits for
27643 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
27644 (vect_gen_while): New function.
27645 * config/aarch64/aarch64.md (umax<mode>3): New expander.
27646 (aarch64_uqdec<mode>): New insn.
27648 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27649 Alan Hayward <alan.hayward@arm.com>
27650 David Sherwood <david.sherwood@arm.com>
27652 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
27653 (reduc_xor_scal_optab): New optabs.
27654 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
27655 (reduc_xor_scal_@var{m}): Document.
27656 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
27657 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
27658 internal functions.
27659 * fold-const-call.c (fold_const_call): Handle them.
27660 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
27661 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
27662 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
27663 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
27664 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
27665 (UNSPEC_XORV): New unspecs.
27666 (optab): Add entries for them.
27667 (BITWISEV): New int iterator.
27668 (bit_reduc_op): New int attributes.
27670 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27671 Alan Hayward <alan.hayward@arm.com>
27672 David Sherwood <david.sherwood@arm.com>
27674 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
27675 * internal-fn.def (VEC_SHL_INSERT): New internal function.
27676 * optabs.def (vec_shl_insert_optab): New optab.
27677 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
27678 (duplicate_and_interleave): Likewise.
27679 * tree-vect-loop.c: Include internal-fn.h.
27680 (neutral_op_for_slp_reduction): New function, split out from
27681 get_initial_defs_for_reduction.
27682 (get_initial_def_for_reduction): Handle option 2 for variable-length
27683 vectors by loading the neutral value into a vector and then shifting
27684 the initial value into element 0.
27685 (get_initial_defs_for_reduction): Replace the code argument with
27686 the neutral value calculated by neutral_op_for_slp_reduction.
27687 Use gimple_build_vector for constant-length vectors.
27688 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
27689 but the first group_size elements have a neutral value.
27690 Use duplicate_and_interleave otherwise.
27691 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
27692 Update call to get_initial_defs_for_reduction. Handle SLP
27693 reductions for variable-length vectors by creating one vector
27694 result for each scalar result, with the elements associated
27695 with other scalar results stubbed out with the neutral value.
27696 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
27697 Require IFN_VEC_SHL_INSERT for double reductions on
27698 variable-length vectors, or SLP reductions that have
27699 a neutral value. Require can_duplicate_and_interleave_p
27700 support for variable-length unchained SLP reductions if there
27701 is no neutral value, such as for MIN/MAX reductions. Also require
27702 the number of vector elements to be a multiple of the number of
27703 SLP statements when doing variable-length unchained SLP reductions.
27704 Update call to vect_create_epilog_for_reduction.
27705 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
27706 and remove initial values.
27707 (duplicate_and_interleave): Make public.
27708 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
27709 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
27711 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27712 Alan Hayward <alan.hayward@arm.com>
27713 David Sherwood <david.sherwood@arm.com>
27715 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
27716 (can_duplicate_and_interleave_p): New function.
27717 (vect_get_and_check_slp_defs): Take the vector of statements
27718 rather than just the current one. Remove excess parentheses.
27719 Restriction rejectinon of vect_constant_def and vect_external_def
27720 for variable-length vectors to boolean types, or types for which
27721 can_duplicate_and_interleave_p is false.
27722 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
27723 (duplicate_and_interleave): New function.
27724 (vect_get_constant_vectors): Use gimple_build_vector for
27725 constant-length vectors and suitable variable-length constant
27726 vectors. Use duplicate_and_interleave for other variable-length
27727 vectors. Don't defer the update when inserting new statements.
27729 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27730 Alan Hayward <alan.hayward@arm.com>
27731 David Sherwood <david.sherwood@arm.com>
27733 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
27734 min_profitable_iters doesn't go negative.
27736 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27737 Alan Hayward <alan.hayward@arm.com>
27738 David Sherwood <david.sherwood@arm.com>
27740 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
27741 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
27742 * optabs.def (vec_mask_load_lanes_optab): New optab.
27743 (vec_mask_store_lanes_optab): Likewise.
27744 * internal-fn.def (MASK_LOAD_LANES): New internal function.
27745 (MASK_STORE_LANES): Likewise.
27746 * internal-fn.c (mask_load_lanes_direct): New macro.
27747 (mask_store_lanes_direct): Likewise.
27748 (expand_mask_load_optab_fn): Handle masked operations.
27749 (expand_mask_load_lanes_optab_fn): New macro.
27750 (expand_mask_store_optab_fn): Handle masked operations.
27751 (expand_mask_store_lanes_optab_fn): New macro.
27752 (direct_mask_load_lanes_optab_supported_p): Likewise.
27753 (direct_mask_store_lanes_optab_supported_p): Likewise.
27754 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
27756 (vect_load_lanes_supported): Likewise.
27757 * tree-vect-data-refs.c (strip_conversion): New function.
27758 (can_group_stmts_p): Likewise.
27759 (vect_analyze_data_ref_accesses): Use it instead of checking
27760 for a pair of assignments.
27761 (vect_store_lanes_supported): Take a masked_p parameter.
27762 (vect_load_lanes_supported): Likewise.
27763 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
27764 vect_store_lanes_supported and vect_load_lanes_supported.
27765 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
27766 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
27767 parameter. Don't allow gaps for masked accesses.
27768 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
27769 and vect_load_lanes_supported.
27770 (get_load_store_type): Take a masked_p parameter and update
27771 call to get_group_load_store_type.
27772 (vectorizable_store): Update call to get_load_store_type.
27773 Handle IFN_MASK_STORE_LANES.
27774 (vectorizable_load): Update call to get_load_store_type.
27775 Handle IFN_MASK_LOAD_LANES.
27777 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27778 Alan Hayward <alan.hayward@arm.com>
27779 David Sherwood <david.sherwood@arm.com>
27781 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
27783 * config/aarch64/aarch64-protos.h
27784 (aarch64_sve_struct_memory_operand_p): Declare.
27785 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
27786 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
27787 (VPRED, vpred): Handle SVE structure modes.
27788 * config/aarch64/constraints.md (Utx): New constraint.
27789 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
27790 (aarch64_sve_struct_nonimmediate_operand): New predicates.
27791 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
27792 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
27793 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
27794 structure modes. Split into pieces after RA.
27795 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
27796 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
27798 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
27799 SVE structure modes.
27800 (aarch64_classify_address): Likewise.
27801 (sizetochar): Move earlier in file.
27802 (aarch64_print_operand): Handle SVE register lists.
27803 (aarch64_array_mode): New function.
27804 (aarch64_sve_struct_memory_operand_p): Likewise.
27805 (TARGET_ARRAY_MODE): Redefine.
27807 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27808 Alan Hayward <alan.hayward@arm.com>
27809 David Sherwood <david.sherwood@arm.com>
27811 * target.def (array_mode): New target hook.
27812 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
27813 * doc/tm.texi: Regenerate.
27814 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
27815 * hooks.c (hook_optmode_mode_uhwi_none): New function.
27816 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
27817 targetm.array_mode.
27818 * stor-layout.c (mode_for_array): Likewise. Support polynomial
27821 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27822 Alan Hayward <alan.hayward@arm.com>
27823 David Sherwood <david.sherwood@arm.com>
27825 * fold-const.c (fold_binary_loc): Check the argument types
27826 rather than the result type when testing for a vector operation.
27828 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27830 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
27831 * doc/tm.texi: Regenerate.
27833 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27834 Alan Hayward <alan.hayward@arm.com>
27835 David Sherwood <david.sherwood@arm.com>
27837 * doc/invoke.texi (-msve-vector-bits=): Document new option.
27838 (sve): Document new AArch64 extension.
27839 * doc/md.texi (w): Extend the description of the AArch64
27840 constraint to include SVE vectors.
27841 (Upl, Upa): Document new AArch64 predicate constraints.
27842 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
27844 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
27845 (msve-vector-bits=): New option.
27846 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
27847 SVE when these are disabled.
27848 (sve): New extension.
27849 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
27850 modes. Adjust their number of units based on aarch64_sve_vg.
27851 (MAX_BITSIZE_MODE_ANY_MODE): Define.
27852 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
27853 aarch64_addr_query_type.
27854 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
27855 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
27856 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
27857 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
27858 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
27859 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
27860 (aarch64_simd_imm_zero_p): Delete.
27861 (aarch64_check_zero_based_sve_index_immediate): Declare.
27862 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
27863 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
27864 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
27865 (aarch64_sve_float_mul_immediate_p): Likewise.
27866 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
27867 rather than an rtx.
27868 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
27869 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
27870 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
27871 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
27872 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
27873 (aarch64_regmode_natural_size): Likewise.
27874 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
27875 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
27877 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
27878 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
27879 for VG and the SVE predicate registers.
27880 (V_ALIASES): Add a "z"-prefixed alias.
27881 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
27882 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
27883 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
27884 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
27885 (REG_CLASS_NAMES): Add entries for them.
27886 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
27887 and the predicate registers.
27888 (aarch64_sve_vg): Declare.
27889 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
27890 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
27891 (REGMODE_NATURAL_SIZE): Define.
27892 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
27894 * config/aarch64/aarch64.c: Include cfgrtl.h.
27895 (simd_immediate_info): Add a constructor for series vectors,
27896 and an associated step field.
27897 (aarch64_sve_vg): New variable.
27898 (aarch64_dbx_register_number): Handle VG and the predicate registers.
27899 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
27900 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
27901 (VEC_ANY_DATA, VEC_STRUCT): New constants.
27902 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
27903 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
27904 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
27905 (aarch64_get_mask_mode): New functions.
27906 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
27907 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
27908 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
27909 predicate modes and predicate registers. Explicitly restrict
27910 GPRs to modes of 16 bytes or smaller. Only allow FP registers
27911 to store a vector mode if it is recognized by
27912 aarch64_classify_vector_mode.
27913 (aarch64_regmode_natural_size): New function.
27914 (aarch64_hard_regno_caller_save_mode): Return the original mode
27916 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
27917 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
27918 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
27919 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
27921 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
27922 does not overlap dest if the function is frame-related. Handle
27924 (aarch64_split_add_offset): New function.
27925 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
27926 them aarch64_add_offset.
27927 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
27928 and update call to aarch64_sub_sp.
27929 (aarch64_add_cfa_expression): New function.
27930 (aarch64_expand_prologue): Pass extra temporary registers to the
27931 functions above. Handle the case in which we need to emit new
27932 DW_CFA_expressions for registers that were originally saved
27933 relative to the stack pointer, but now have to be expressed
27934 relative to the frame pointer.
27935 (aarch64_output_mi_thunk): Pass extra temporary registers to the
27937 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
27938 IP0 and IP1 values for SVE frames.
27939 (aarch64_expand_vec_series): New function.
27940 (aarch64_expand_sve_widened_duplicate): Likewise.
27941 (aarch64_expand_sve_const_vector): Likewise.
27942 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
27943 Handle SVE constants. Use emit_move_insn to move a force_const_mem
27944 into the register, rather than emitting a SET directly.
27945 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
27946 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
27947 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
27948 (offset_9bit_signed_scaled_p): New functions.
27949 (aarch64_replicate_bitmask_imm): New function.
27950 (aarch64_bitmask_imm): Use it.
27951 (aarch64_cannot_force_const_mem): Reject expressions involving
27952 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
27953 (aarch64_classify_index): Handle SVE indices, by requiring
27954 a plain register index with a scale that matches the element size.
27955 (aarch64_classify_address): Handle SVE addresses. Assert that
27956 the mode of the address is VOIDmode or an integer mode.
27957 Update call to aarch64_classify_symbol.
27958 (aarch64_classify_symbolic_expression): Update call to
27959 aarch64_classify_symbol.
27960 (aarch64_const_vec_all_in_range_p): New function.
27961 (aarch64_print_vector_float_operand): Likewise.
27962 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
27963 "vN" for FP registers with SVE modes. Handle (const ...) vectors
27964 and the FP immediates 1.0 and 0.5.
27965 (aarch64_print_address_internal): Handle SVE addresses.
27966 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
27967 (aarch64_regno_regclass): Handle predicate registers.
27968 (aarch64_secondary_reload): Handle big-endian reloads of SVE
27970 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
27971 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
27972 (aarch64_convert_sve_vector_bits): New function.
27973 (aarch64_override_options): Use it to handle -msve-vector-bits=.
27974 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
27975 rather than an rtx.
27976 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
27977 Handle SVE vector and predicate modes. Accept VL-based constants
27978 that need only one temporary register, and VL offsets that require
27979 no temporary registers.
27980 (aarch64_conditional_register_usage): Mark the predicate registers
27981 as fixed if SVE isn't available.
27982 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
27983 Return true for SVE vector and predicate modes.
27984 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
27985 rather than an unsigned int. Handle SVE modes.
27986 (aarch64_preferred_simd_mode): Update call accordingly. Handle
27988 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
27990 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
27991 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
27992 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
27993 (aarch64_sve_float_mul_immediate_p): New functions.
27994 (aarch64_sve_valid_immediate): New function.
27995 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
27996 Explicitly reject structure modes. Check for INDEX constants.
27997 Handle PTRUE and PFALSE constants.
27998 (aarch64_check_zero_based_sve_index_immediate): New function.
27999 (aarch64_simd_imm_zero_p): Delete.
28000 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
28001 vector modes. Accept constants in the range of CNT[BHWD].
28002 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
28003 ask for an Advanced SIMD mode.
28004 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
28005 (aarch64_simd_vector_alignment): Handle SVE predicates.
28006 (aarch64_vectorize_preferred_vector_alignment): New function.
28007 (aarch64_simd_vector_alignment_reachable): Use it instead of
28009 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
28010 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
28012 (MAX_VECT_LEN): Delete.
28013 (expand_vec_perm_d): Add a vec_flags field.
28014 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
28015 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
28016 (aarch64_evpc_ext): Don't apply a big-endian lane correction
28018 (aarch64_evpc_rev): Rename to...
28019 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
28020 (aarch64_evpc_rev_global): New function.
28021 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
28022 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
28024 (aarch64_evpc_sve_tbl): New function.
28025 (aarch64_expand_vec_perm_const_1): Update after rename of
28026 aarch64_evpc_rev. Handle SVE permutes too, trying
28027 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
28028 than aarch64_evpc_tbl.
28029 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
28030 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
28031 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
28032 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
28033 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
28034 (aarch64_expand_sve_vcond): New functions.
28035 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
28036 of aarch64_vector_mode_p.
28037 (aarch64_dwarf_poly_indeterminate_value): New function.
28038 (aarch64_compute_pressure_classes): Likewise.
28039 (aarch64_can_change_mode_class): Likewise.
28040 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
28041 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
28042 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
28043 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
28044 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
28045 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
28046 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
28047 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
28049 (Dn, Dl, Dr): Accept const as well as const_vector.
28050 (Dz): Likewise. Compare against CONST0_RTX.
28051 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
28052 of "vector" where appropriate.
28053 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
28054 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
28055 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
28056 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
28057 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
28058 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
28059 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
28060 (v_int_equiv): Extend to SVE modes.
28061 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
28063 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
28064 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
28065 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
28066 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
28067 (SVE_COND_FP_CMP): New int iterators.
28068 (perm_hilo): Handle the new unpack unspecs.
28069 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
28071 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
28072 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
28073 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
28074 (aarch64_equality_operator, aarch64_constant_vector_operand)
28075 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
28076 (aarch64_sve_nonimmediate_operand): Likewise.
28077 (aarch64_sve_general_operand): Likewise.
28078 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
28079 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
28080 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
28081 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
28082 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
28083 (aarch64_sve_float_arith_immediate): Likewise.
28084 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
28085 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
28086 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
28087 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
28088 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
28089 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
28090 (aarch64_sve_float_arith_operand): Likewise.
28091 (aarch64_sve_float_arith_with_sub_operand): Likewise.
28092 (aarch64_sve_float_mul_operand): Likewise.
28093 (aarch64_sve_vec_perm_operand): Likewise.
28094 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
28095 (aarch64_mov_operand): Accept const_poly_int and const_vector.
28096 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
28097 as well as const_vector.
28098 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
28099 in file. Use CONST0_RTX and CONSTM1_RTX.
28100 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
28101 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
28102 Use aarch64_simd_imm_zero.
28103 * config/aarch64/aarch64-sve.md: New file.
28104 * config/aarch64/aarch64.md: Include it.
28105 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
28106 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
28107 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
28108 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
28109 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
28110 (sve): New attribute.
28111 (enabled): Disable instructions with the sve attribute unless
28113 (movqi, movhi): Pass CONST_POLY_INT operaneds through
28114 aarch64_expand_mov_immediate.
28115 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
28116 CNT[BHSD] immediates.
28117 (movti): Split CONST_POLY_INT moves into two halves.
28118 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
28119 Split additions that need a temporary here if the destination
28120 is the stack pointer.
28121 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
28122 (*add<mode>3_poly_1): New instruction.
28123 (set_clobber_cc): New expander.
28125 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28127 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
28128 parameter and use it instead of GET_MODE_SIZE (innermode). Use
28129 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
28130 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
28131 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
28132 Change innermode from fixed_mode_size to machine_mode.
28133 (simplify_subreg): Update call accordingly. Handle a constant-sized
28134 subreg of a variable-length CONST_VECTOR.
28136 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28137 Alan Hayward <alan.hayward@arm.com>
28138 David Sherwood <david.sherwood@arm.com>
28140 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
28141 (add_offset_to_base): New function, split out from...
28142 (create_mem_ref): ...here. When handling a scale other than 1,
28143 check first whether the address is valid without the offset.
28144 Add it into the base if so, leaving the index and scale as-is.
28146 2018-01-12 Jakub Jelinek <jakub@redhat.com>
28149 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
28150 fold_for_warn before checking if arg2 is INTEGER_CST.
28152 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
28154 * config/rs6000/predicates.md (load_multiple_operation): Delete.
28155 (store_multiple_operation): Delete.
28156 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
28157 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
28158 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
28159 guarded by TARGET_STRING.
28160 (rs6000_output_load_multiple): Delete.
28161 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
28162 OPTION_MASK_STRING / TARGET_STRING handling.
28163 (print_operand) <'N', 'O'>: Add comment that these are unused now.
28164 (const rs6000_opt_masks) <"string">: Change mask to 0.
28165 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
28166 (MASK_STRING): Delete.
28167 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
28169 (load_multiple): Delete.
28176 (store_multiple): Delete.
28183 (movmemsi_8reg): Delete.
28184 (corresponding unnamed define_insn): Delete.
28185 (movmemsi_6reg): Delete.
28186 (corresponding unnamed define_insn): Delete.
28187 (movmemsi_4reg): Delete.
28188 (corresponding unnamed define_insn): Delete.
28189 (movmemsi_2reg): Delete.
28190 (corresponding unnamed define_insn): Delete.
28191 (movmemsi_1reg): Delete.
28192 (corresponding unnamed define_insn): Delete.
28193 * config/rs6000/rs6000.opt (mno-string): New.
28194 (mstring): Replace by deprecation warning stub.
28195 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
28197 2018-01-12 Jakub Jelinek <jakub@redhat.com>
28199 * regrename.c (regrename_do_replace): If replacing the same
28200 reg multiple times, try to reuse last created gen_raw_REG.
28203 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
28204 main to workaround a bug in GDB.
28206 2018-01-12 Tom de Vries <tom@codesourcery.com>
28209 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
28211 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
28213 PR rtl-optimization/80481
28214 * ira-color.c (get_cap_member): New function.
28215 (allocnos_conflict_by_live_ranges_p): Use it.
28216 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
28217 (setup_slot_coalesced_allocno_live_ranges): Ditto.
28219 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
28222 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
28223 (*saddl_se_1): Ditto.
28224 (*ssubsi_1): Ditto.
28225 (*ssubl_se_1): Ditto.
28227 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
28229 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
28230 rather than wi::to_widest for DR_INITs.
28231 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
28232 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
28233 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
28235 (vect_analyze_group_access_1): Note that here.
28237 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
28239 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
28240 polynomial type sizes.
28242 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
28244 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
28245 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
28246 (gimple_add_tmp_var): Likewise.
28248 2018-01-12 Martin Liska <mliska@suse.cz>
28250 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
28251 (gimple_alloc_sizes): Likewise.
28252 (dump_gimple_statistics): Use PRIu64 in printf format.
28253 * gimple.h: Change uint64_t to int.
28255 2018-01-12 Martin Liska <mliska@suse.cz>
28257 * tree-core.h: Use uint64_t instead of int.
28258 * tree.c (tree_node_counts): Likewise.
28259 (tree_node_sizes): Likewise.
28260 (dump_tree_statistics): Use PRIu64 in printf format.
28262 2018-01-12 Martin Liska <mliska@suse.cz>
28264 * Makefile.in: As qsort_chk is implemented in vec.c, add
28265 vec.o to linkage of gencfn-macros.
28266 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
28267 passing the info to record_node_allocation_statistics.
28268 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
28270 * ggc-common.c (struct ggc_usage): Add operator== and use
28271 it in operator< and compare function.
28272 * mem-stats.h (struct mem_usage): Likewise.
28273 * vec.c (struct vec_usage): Remove operator< and compare
28274 function. Can be simply inherited.
28276 2018-01-12 Martin Jambor <mjambor@suse.cz>
28279 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
28280 * tree-ssa-math-opts.c: Include domwalk.h.
28281 (convert_mult_to_fma_1): New function.
28282 (fma_transformation_info): New type.
28283 (fma_deferring_state): Likewise.
28284 (cancel_fma_deferring): New function.
28285 (result_of_phi): Likewise.
28286 (last_fma_candidate_feeds_initial_phi): Likewise.
28287 (convert_mult_to_fma): Added deferring logic, split actual
28288 transformation to convert_mult_to_fma_1.
28289 (math_opts_dom_walker): New type.
28290 (math_opts_dom_walker::after_dom_children): New method, body moved
28291 here from pass_optimize_widening_mul::execute, added deferring logic
28293 (pass_optimize_widening_mul::execute): Moved most of code to
28294 math_opts_dom_walker::after_dom_children.
28295 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
28296 * config/i386/i386.c (ix86_option_override_internal): Added
28297 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
28299 2018-01-12 Richard Biener <rguenther@suse.de>
28302 * dwarf2out.c (gen_variable_die): Do not reset old_die for
28303 inline instance vars.
28305 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
28308 * config/rx/rx.c (rx_is_restricted_memory_address):
28309 Handle SUBREG case.
28311 2018-01-12 Richard Biener <rguenther@suse.de>
28313 PR tree-optimization/80846
28314 * target.def (split_reduction): New target hook.
28315 * targhooks.c (default_split_reduction): New function.
28316 * targhooks.h (default_split_reduction): Declare.
28317 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
28318 target requests first reduce vectors by combining low and high
28320 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
28321 (get_vectype_for_scalar_type_and_size): Export.
28322 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
28323 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
28324 * doc/tm.texi: Regenerate.
28325 * config/i386/i386.c (ix86_split_reduction): Implement
28326 TARGET_VECTORIZE_SPLIT_REDUCTION.
28328 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
28331 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
28332 in PIC mode except for TARGET_VXWORKS_RTP.
28333 * config/sparc/sparc.c: Include cfgrtl.h.
28334 (TARGET_INIT_PIC_REG): Define.
28335 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
28336 (sparc_pic_register_p): New predicate.
28337 (sparc_legitimate_address_p): Use it.
28338 (sparc_legitimize_pic_address): Likewise.
28339 (sparc_delegitimize_address): Likewise.
28340 (sparc_mode_dependent_address_p): Likewise.
28341 (gen_load_pcrel_sym): Remove 4th parameter.
28342 (load_got_register): Adjust call to above. Remove obsolete stuff.
28343 (sparc_expand_prologue): Do not call load_got_register here.
28344 (sparc_flat_expand_prologue): Likewise.
28345 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
28346 (sparc_use_pseudo_pic_reg): New function.
28347 (sparc_init_pic_reg): Likewise.
28348 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
28349 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
28351 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
28353 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28354 Add item for branch_cost.
28356 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
28358 PR rtl-optimization/83565
28359 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
28360 not extend the result to a larger mode for rotate operations.
28361 (num_sign_bit_copies1): Likewise.
28363 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
28366 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
28368 Use values-Xc.o for -pedantic.
28369 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
28371 2018-01-12 Martin Liska <mliska@suse.cz>
28374 * ipa-devirt.c (final_warning_record::grow_type_warnings):
28376 (possible_polymorphic_call_targets): Use it.
28377 (ipa_devirt): Likewise.
28379 2018-01-12 Martin Liska <mliska@suse.cz>
28381 * profile-count.h (enum profile_quality): Use 0 as invalid
28382 enum value of profile_quality.
28384 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
28386 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
28387 -mext-string options.
28389 2018-01-12 Richard Biener <rguenther@suse.de>
28391 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
28392 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
28393 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
28395 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
28397 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
28399 * configure.ac (--with-long-double-format): Add support for the
28400 configuration option to change the default long double format on
28402 * config.gcc (powerpc*-linux*-*): Likewise.
28403 * configure: Regenerate.
28404 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
28405 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
28406 used without modification.
28408 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
28410 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
28411 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
28412 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
28413 MISC_BUILTIN_SPEC_BARRIER.
28414 (rs6000_init_builtins): Likewise.
28415 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
28417 (speculation_barrier): New define_insn.
28418 * doc/extend.texi: Document __builtin_speculation_barrier.
28420 2018-01-11 Jakub Jelinek <jakub@redhat.com>
28423 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
28424 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
28425 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
28427 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
28428 integral modes instead of "ss" and "sd".
28429 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
28430 vectors with 32-bit and 64-bit elements.
28431 (vecdupssescalarmodesuffix): New mode attribute.
28432 (vec_dup<mode>): Use it.
28434 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
28437 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
28438 frame if argument is passed on stack.
28440 2018-01-11 Jakub Jelinek <jakub@redhat.com>
28443 * ree.c (combine_reaching_defs): Optimize also
28444 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
28445 reg2=any_extend(exp); reg1=reg2;, formatting fix.
28447 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
28449 PR middle-end/83189
28450 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
28452 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
28454 PR middle-end/83718
28455 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
28456 after they are computed.
28458 2018-01-11 Bin Cheng <bin.cheng@arm.com>
28460 PR tree-optimization/83695
28461 * gimple-loop-linterchange.cc
28462 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
28463 reset cached scev information after interchange.
28464 (pass_linterchange::execute): Remove call to scev_reset_htab.
28466 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28468 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
28469 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
28470 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
28471 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
28472 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
28473 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
28474 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
28475 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
28476 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
28477 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
28478 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
28479 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
28480 (V_lane_reg): Likewise.
28481 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
28483 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
28484 (vfmal_lane_low<mode>_intrinsic,
28485 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
28486 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
28487 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
28488 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
28489 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
28490 vfmsl_lane_high<mode>_intrinsic): New define_insns.
28492 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28494 * config/arm/arm-cpus.in (fp16fml): New feature.
28495 (ALL_SIMD): Add fp16fml.
28496 (armv8.2-a): Add fp16fml as an option.
28497 (armv8.3-a): Likewise.
28498 (armv8.4-a): Add fp16fml as part of fp16.
28499 * config/arm/arm.h (TARGET_FP16FML): Define.
28500 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
28502 * config/arm/arm-modes.def (V2HF): Define.
28503 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
28504 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
28505 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
28506 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
28507 vfmsl_low, vfmsl_high): New set of builtins.
28508 * config/arm/iterators.md (PLUSMINUS): New code iterator.
28509 (vfml_op): New code attribute.
28510 (VFMLHALVES): New int iterator.
28511 (VFML, VFMLSEL): New mode attributes.
28512 (V_reg): Define mapping for V2HF.
28513 (V_hi, V_lo): New mode attributes.
28514 (VF_constraint): Likewise.
28515 (vfml_half, vfml_half_selector): New int attributes.
28516 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
28518 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
28519 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
28521 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
28522 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
28523 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
28524 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
28526 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
28527 Document new effective target and option set.
28529 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28531 * config/arm/arm-cpus.in (armv8_4): New feature.
28532 (ARMv8_4a): New fgroup.
28533 (armv8.4-a): New arch.
28534 * config/arm/arm-tables.opt: Regenerate.
28535 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
28536 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
28537 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
28538 Add matching rules for -march=armv8.4-a and extensions.
28539 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
28541 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
28544 * config/rx/rx.md (BW): New mode attribute.
28545 (sync_lock_test_and_setsi): Add mode suffix to insn output.
28547 2018-01-11 Richard Biener <rguenther@suse.de>
28549 PR tree-optimization/83435
28550 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
28551 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
28552 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
28554 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
28555 Alan Hayward <alan.hayward@arm.com>
28556 David Sherwood <david.sherwood@arm.com>
28558 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
28560 (aarch64_classify_address): Initialize it. Track polynomial offsets.
28561 (aarch64_print_address_internal): Use it to check for a zero offset.
28563 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
28564 Alan Hayward <alan.hayward@arm.com>
28565 David Sherwood <david.sherwood@arm.com>
28567 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
28568 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
28569 Return a poly_int64 rather than a HOST_WIDE_INT.
28570 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
28571 rather than a HOST_WIDE_INT.
28572 * config/aarch64/aarch64.h (aarch64_frame): Protect with
28573 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
28574 hard_fp_offset, frame_size, initial_adjust, callee_offset and
28575 final_offset from HOST_WIDE_INT to poly_int64.
28576 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
28577 to_constant when getting the number of units in an Advanced SIMD
28579 (aarch64_builtin_vectorized_function): Check for a constant number
28581 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
28583 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
28584 attribute instead of GET_MODE_NUNITS.
28585 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
28586 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
28587 GET_MODE_SIZE for fixed-size registers.
28588 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
28589 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
28590 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
28591 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
28592 (aarch64_print_operand, aarch64_print_address_internal)
28593 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
28594 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
28595 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
28596 Handle polynomial GET_MODE_SIZE.
28597 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
28598 wider than SImode without modification.
28599 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
28600 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
28601 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
28602 passing and returning SVE modes.
28603 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
28604 rather than GEN_INT.
28605 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
28606 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
28607 (aarch64_allocate_and_probe_stack_space): Likewise.
28608 (aarch64_layout_frame): Cope with polynomial offsets.
28609 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
28610 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
28611 polynomial offsets.
28612 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
28613 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
28614 poly_int64 rather than a HOST_WIDE_INT.
28615 (aarch64_get_separate_components, aarch64_process_components)
28616 (aarch64_expand_prologue, aarch64_expand_epilogue)
28617 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
28618 (aarch64_anchor_offset): New function, split out from...
28619 (aarch64_legitimize_address): ...here.
28620 (aarch64_builtin_vectorization_cost): Handle polynomial
28621 TYPE_VECTOR_SUBPARTS.
28622 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
28624 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
28625 number of elements from the PARALLEL rather than the mode.
28626 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
28627 rather than GET_MODE_BITSIZE.
28628 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
28629 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
28630 (aarch64_expand_vec_perm_const_1): Handle polynomial
28631 d->perm.length () and d->perm elements.
28632 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
28633 Apply to_constant to d->perm elements.
28634 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
28635 polynomial CONST_VECTOR_NUNITS.
28636 (aarch64_move_pointer): Take amount as a poly_int64 rather
28638 (aarch64_progress_pointer): Avoid temporary variable.
28639 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
28640 the mode attribute instead of GET_MODE.
28642 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
28643 Alan Hayward <alan.hayward@arm.com>
28644 David Sherwood <david.sherwood@arm.com>
28646 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
28647 x exists before using it.
28648 (aarch64_add_constant_internal): Rename to...
28649 (aarch64_add_offset_1): ...this. Replace regnum with separate
28650 src and dest rtxes. Handle the case in which they're different,
28651 including when the offset is zero. Replace scratchreg with an rtx.
28652 Use 2 additions if there is no spare register into which we can
28653 move a 16-bit constant.
28654 (aarch64_add_constant): Delete.
28655 (aarch64_add_offset): Replace reg with separate src and dest
28656 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
28657 Use aarch64_add_offset_1.
28658 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
28659 an rtx rather than an int. Take the delta as a poly_int64
28660 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
28661 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
28662 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
28663 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
28664 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
28665 and aarch64_add_sp.
28666 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
28667 aarch64_add_constant.
28669 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
28671 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
28672 Use scalar_float_mode.
28674 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
28676 * config/aarch64/aarch64-simd.md
28677 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
28678 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
28679 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
28680 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
28681 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
28682 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
28683 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
28684 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
28685 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
28686 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
28688 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
28691 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
28692 targ_options->x_arm_arch_string is non NULL.
28694 2018-01-11 Tamar Christina <tamar.christina@arm.com>
28696 * config/aarch64/aarch64.h
28697 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
28699 2018-01-11 Sudakshina Das <sudi.das@arm.com>
28702 * expmed.c (emit_store_flag_force): Swap if const op0
28703 and change VOIDmode to mode of op0.
28705 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
28707 PR rtl-optimization/83761
28708 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
28709 than bytes to mode_for_size.
28711 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
28713 PR middle-end/83189
28714 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
28715 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
28718 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
28720 PR middle-end/83575
28721 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
28722 when in layout mode.
28723 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
28724 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
28727 2018-01-10 Michael Collison <michael.collison@arm.com>
28729 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
28730 * config/aarch64/aarch64-option-extension.def: Add
28731 AARCH64_OPT_EXTENSION of 'fp16fml'.
28732 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
28733 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
28734 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
28735 * config/aarch64/constraints.md (Ui7): New constraint.
28736 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
28737 (VFMLA_SEL_W): Ditto.
28740 (VFMLA16_LOW): New int iterator.
28741 (VFMLA16_HIGH): Ditto.
28742 (UNSPEC_FMLAL): New unspec.
28743 (UNSPEC_FMLSL): Ditto.
28744 (UNSPEC_FMLAL2): Ditto.
28745 (UNSPEC_FMLSL2): Ditto.
28746 (f16mac): New code attribute.
28747 * config/aarch64/aarch64-simd-builtins.def
28748 (aarch64_fmlal_lowv2sf): Ditto.
28749 (aarch64_fmlsl_lowv2sf): Ditto.
28750 (aarch64_fmlalq_lowv4sf): Ditto.
28751 (aarch64_fmlslq_lowv4sf): Ditto.
28752 (aarch64_fmlal_highv2sf): Ditto.
28753 (aarch64_fmlsl_highv2sf): Ditto.
28754 (aarch64_fmlalq_highv4sf): Ditto.
28755 (aarch64_fmlslq_highv4sf): Ditto.
28756 (aarch64_fmlal_lane_lowv2sf): Ditto.
28757 (aarch64_fmlsl_lane_lowv2sf): Ditto.
28758 (aarch64_fmlal_laneq_lowv2sf): Ditto.
28759 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
28760 (aarch64_fmlalq_lane_lowv4sf): Ditto.
28761 (aarch64_fmlsl_lane_lowv4sf): Ditto.
28762 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
28763 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
28764 (aarch64_fmlal_lane_highv2sf): Ditto.
28765 (aarch64_fmlsl_lane_highv2sf): Ditto.
28766 (aarch64_fmlal_laneq_highv2sf): Ditto.
28767 (aarch64_fmlsl_laneq_highv2sf): Ditto.
28768 (aarch64_fmlalq_lane_highv4sf): Ditto.
28769 (aarch64_fmlsl_lane_highv4sf): Ditto.
28770 (aarch64_fmlalq_laneq_highv4sf): Ditto.
28771 (aarch64_fmlsl_laneq_highv4sf): Ditto.
28772 * config/aarch64/aarch64-simd.md:
28773 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
28774 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
28775 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
28776 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
28777 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
28778 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
28779 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
28780 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
28781 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
28782 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
28783 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
28784 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
28785 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
28786 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
28787 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
28788 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
28789 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
28790 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
28791 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
28792 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
28793 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
28794 (vfmlsl_low_u32): Ditto.
28795 (vfmlalq_low_u32): Ditto.
28796 (vfmlslq_low_u32): Ditto.
28797 (vfmlal_high_u32): Ditto.
28798 (vfmlsl_high_u32): Ditto.
28799 (vfmlalq_high_u32): Ditto.
28800 (vfmlslq_high_u32): Ditto.
28801 (vfmlal_lane_low_u32): Ditto.
28802 (vfmlsl_lane_low_u32): Ditto.
28803 (vfmlal_laneq_low_u32): Ditto.
28804 (vfmlsl_laneq_low_u32): Ditto.
28805 (vfmlalq_lane_low_u32): Ditto.
28806 (vfmlslq_lane_low_u32): Ditto.
28807 (vfmlalq_laneq_low_u32): Ditto.
28808 (vfmlslq_laneq_low_u32): Ditto.
28809 (vfmlal_lane_high_u32): Ditto.
28810 (vfmlsl_lane_high_u32): Ditto.
28811 (vfmlal_laneq_high_u32): Ditto.
28812 (vfmlsl_laneq_high_u32): Ditto.
28813 (vfmlalq_lane_high_u32): Ditto.
28814 (vfmlslq_lane_high_u32): Ditto.
28815 (vfmlalq_laneq_high_u32): Ditto.
28816 (vfmlslq_laneq_high_u32): Ditto.
28817 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
28818 (AARCH64_FL_FOR_ARCH8_4): New.
28819 (AARCH64_ISA_F16FML): New ISA flag.
28820 (TARGET_F16FML): New feature flag for fp16fml.
28821 (doc/invoke.texi): Document new fp16fml option.
28823 2018-01-10 Michael Collison <michael.collison@arm.com>
28825 * config/aarch64/aarch64-builtins.c:
28826 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
28827 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
28828 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
28829 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
28830 (AARCH64_ISA_SHA3): New ISA flag.
28831 (TARGET_SHA3): New feature flag for sha3.
28832 * config/aarch64/iterators.md (sha512_op): New int attribute.
28833 (CRYPTO_SHA512): New int iterator.
28834 (UNSPEC_SHA512H): New unspec.
28835 (UNSPEC_SHA512H2): Ditto.
28836 (UNSPEC_SHA512SU0): Ditto.
28837 (UNSPEC_SHA512SU1): Ditto.
28838 * config/aarch64/aarch64-simd-builtins.def
28839 (aarch64_crypto_sha512hqv2di): New builtin.
28840 (aarch64_crypto_sha512h2qv2di): Ditto.
28841 (aarch64_crypto_sha512su0qv2di): Ditto.
28842 (aarch64_crypto_sha512su1qv2di): Ditto.
28843 (aarch64_eor3qv8hi): Ditto.
28844 (aarch64_rax1qv2di): Ditto.
28845 (aarch64_xarqv2di): Ditto.
28846 (aarch64_bcaxqv8hi): Ditto.
28847 * config/aarch64/aarch64-simd.md:
28848 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
28849 (aarch64_crypto_sha512su0qv2di): Ditto.
28850 (aarch64_crypto_sha512su1qv2di): Ditto.
28851 (aarch64_eor3qv8hi): Ditto.
28852 (aarch64_rax1qv2di): Ditto.
28853 (aarch64_xarqv2di): Ditto.
28854 (aarch64_bcaxqv8hi): Ditto.
28855 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
28856 (vsha512h2q_u64): Ditto.
28857 (vsha512su0q_u64): Ditto.
28858 (vsha512su1q_u64): Ditto.
28859 (veor3q_u16): Ditto.
28860 (vrax1q_u64): Ditto.
28861 (vxarq_u64): Ditto.
28862 (vbcaxq_u16): Ditto.
28863 * config/arm/types.md (crypto_sha512): New type attribute.
28864 (crypto_sha3): Ditto.
28865 (doc/invoke.texi): Document new sha3 option.
28867 2018-01-10 Michael Collison <michael.collison@arm.com>
28869 * config/aarch64/aarch64-builtins.c:
28870 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
28871 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
28872 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
28873 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
28874 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
28875 (AARCH64_ISA_SM4): New ISA flag.
28876 (TARGET_SM4): New feature flag for sm4.
28877 * config/aarch64/aarch64-simd-builtins.def
28878 (aarch64_sm3ss1qv4si): Ditto.
28879 (aarch64_sm3tt1aq4si): Ditto.
28880 (aarch64_sm3tt1bq4si): Ditto.
28881 (aarch64_sm3tt2aq4si): Ditto.
28882 (aarch64_sm3tt2bq4si): Ditto.
28883 (aarch64_sm3partw1qv4si): Ditto.
28884 (aarch64_sm3partw2qv4si): Ditto.
28885 (aarch64_sm4eqv4si): Ditto.
28886 (aarch64_sm4ekeyqv4si): Ditto.
28887 * config/aarch64/aarch64-simd.md:
28888 (aarch64_sm3ss1qv4si): Ditto.
28889 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
28890 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
28891 (aarch64_sm4eqv4si): Ditto.
28892 (aarch64_sm4ekeyqv4si): Ditto.
28893 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
28894 (sm3part_op): Ditto.
28895 (CRYPTO_SM3TT): Ditto.
28896 (CRYPTO_SM3PART): Ditto.
28897 (UNSPEC_SM3SS1): New unspec.
28898 (UNSPEC_SM3TT1A): Ditto.
28899 (UNSPEC_SM3TT1B): Ditto.
28900 (UNSPEC_SM3TT2A): Ditto.
28901 (UNSPEC_SM3TT2B): Ditto.
28902 (UNSPEC_SM3PARTW1): Ditto.
28903 (UNSPEC_SM3PARTW2): Ditto.
28904 (UNSPEC_SM4E): Ditto.
28905 (UNSPEC_SM4EKEY): Ditto.
28906 * config/aarch64/constraints.md (Ui2): New constraint.
28907 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
28908 * config/arm/types.md (crypto_sm3): New type attribute.
28909 (crypto_sm4): Ditto.
28910 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
28911 (vsm3tt1aq_u32): Ditto.
28912 (vsm3tt1bq_u32): Ditto.
28913 (vsm3tt2aq_u32): Ditto.
28914 (vsm3tt2bq_u32): Ditto.
28915 (vsm3partw1q_u32): Ditto.
28916 (vsm3partw2q_u32): Ditto.
28917 (vsm4eq_u32): Ditto.
28918 (vsm4ekeyq_u32): Ditto.
28919 (doc/invoke.texi): Document new sm4 option.
28921 2018-01-10 Michael Collison <michael.collison@arm.com>
28923 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
28924 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
28925 (AARCH64_FL_FOR_ARCH8_4): New.
28926 (AARCH64_FL_V8_4): New flag.
28927 (doc/invoke.texi): Document new armv8.4-a option.
28929 2018-01-10 Michael Collison <michael.collison@arm.com>
28931 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
28932 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
28933 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
28934 * config/aarch64/aarch64-option-extension.def: Add
28935 AARCH64_OPT_EXTENSION of 'sha2'.
28936 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
28937 (crypto): Disable sha2 and aes if crypto disabled.
28938 (crypto): Enable aes and sha2 if enabled.
28939 (simd): Disable sha2 and aes if simd disabled.
28940 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
28942 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
28943 (TARGET_SHA2): New feature flag for sha2.
28944 (TARGET_AES): New feature flag for aes.
28945 * config/aarch64/aarch64-simd.md:
28946 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
28947 conditional on TARGET_AES.
28948 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
28949 (aarch64_crypto_sha1hsi): Make pattern conditional
28951 (aarch64_crypto_sha1hv4si): Ditto.
28952 (aarch64_be_crypto_sha1hv4si): Ditto.
28953 (aarch64_crypto_sha1su1v4si): Ditto.
28954 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
28955 (aarch64_crypto_sha1su0v4si): Ditto.
28956 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
28957 (aarch64_crypto_sha256su0v4si): Ditto.
28958 (aarch64_crypto_sha256su1v4si): Ditto.
28959 (doc/invoke.texi): Document new aes and sha2 options.
28961 2018-01-10 Martin Sebor <msebor@redhat.com>
28963 PR tree-optimization/83781
28964 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
28967 2018-01-11 Martin Sebor <msebor@gmail.com>
28968 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
28970 PR tree-optimization/83501
28971 PR tree-optimization/81703
28973 * tree-ssa-strlen.c (get_string_cst): Rename...
28974 (get_string_len): ...to this. Handle global constants.
28975 (handle_char_store): Adjust.
28977 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
28978 Jim Wilson <jimw@sifive.com>
28980 * config/riscv/riscv-protos.h (riscv_output_return): New.
28981 * config/riscv/riscv.c (struct machine_function): New naked_p field.
28982 (riscv_attribute_table, riscv_output_return),
28983 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
28984 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
28985 (riscv_compute_frame_info): Only compute frame->mask if not a naked
28987 (riscv_expand_prologue): Add early return for naked function.
28988 (riscv_expand_epilogue): Likewise.
28989 (riscv_function_ok_for_sibcall): Return false for naked function.
28990 (riscv_set_current_function): New.
28991 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
28992 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
28993 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
28994 * doc/extend.texi (RISC-V Function Attributes): New.
28996 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
28998 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
28999 check for 128-bit long double before checking TCmode.
29000 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
29001 128-bit long doubles before checking TFmode or TCmode.
29002 (FLOAT128_IBM_P): Likewise.
29004 2018-01-10 Martin Sebor <msebor@redhat.com>
29006 PR tree-optimization/83671
29007 * builtins.c (c_strlen): Unconditionally return zero for the empty
29009 Use -Warray-bounds for warnings.
29010 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
29011 for non-constant array indices with COMPONENT_REF, arrays of
29012 arrays, and pointers to arrays.
29013 (gimple_fold_builtin_strlen): Determine and set length range for
29014 non-constant character arrays.
29016 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
29018 PR middle-end/81897
29019 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
29022 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
29024 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
29026 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
29029 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
29030 VECTOR_MEM_ALTIVEC_OR_VSX_P.
29031 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
29032 indexed_or_indirect_operand predicate.
29033 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
29034 (*vsx_le_perm_load_v8hi): Likewise.
29035 (*vsx_le_perm_load_v16qi): Likewise.
29036 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
29037 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
29038 (*vsx_le_perm_store_v8hi): Likewise.
29039 (*vsx_le_perm_store_v16qi): Likewise.
29040 (eight unnamed splitters): Likewise.
29042 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
29044 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
29045 * config/rs6000/emmintrin.h: Likewise.
29046 * config/rs6000/mmintrin.h: Likewise.
29047 * config/rs6000/xmmintrin.h: Likewise.
29049 2018-01-10 David Malcolm <dmalcolm@redhat.com>
29052 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
29054 * tree.c (tree_nop_conversion): Return true for location wrapper
29056 (maybe_wrap_with_location): New function.
29057 (selftest::check_strip_nops): New function.
29058 (selftest::test_location_wrappers): New function.
29059 (selftest::tree_c_tests): Call it.
29060 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
29061 (maybe_wrap_with_location): New decl.
29062 (EXPR_LOCATION_WRAPPER_P): New macro.
29063 (location_wrapper_p): New inline function.
29064 (tree_strip_any_location_wrapper): New inline function.
29066 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
29069 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
29070 stack_realign_offset for the largest alignment of stack slot
29072 (ix86_find_max_used_stack_alignment): New function.
29073 (ix86_finalize_stack_frame_flags): Use it. Set
29074 max_used_stack_alignment if we don't realign stack.
29075 * config/i386/i386.h (machine_function): Add
29076 max_used_stack_alignment.
29078 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
29080 * config/arm/arm.opt (-mbranch-cost): New option.
29081 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
29084 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
29087 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
29088 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
29090 2018-01-10 Richard Biener <rguenther@suse.de>
29093 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
29094 early out so it also covers the case where we have a non-NULL
29097 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
29099 PR tree-optimization/83753
29100 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
29101 for non-strided grouped accesses if the number of elements is 1.
29103 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
29106 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
29107 * i386.h (TARGET_USE_GATHER): Define.
29108 * x86-tune.def (X86_TUNE_USE_GATHER): New.
29110 2018-01-10 Martin Liska <mliska@suse.cz>
29113 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
29114 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
29116 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
29117 CLEANUP_NO_PARTITIONING is not set.
29119 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
29121 * doc/rtl.texi: Remove documentation of (const ...) wrappers
29122 for vectors, as a partial revert of r254296.
29123 * rtl.h (const_vec_p): Delete.
29124 (const_vec_duplicate_p): Don't test for vector CONSTs.
29125 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
29126 * expmed.c (make_tree): Likewise.
29129 * common.md (E, F): Use CONSTANT_P instead of checking for
29131 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
29132 checking for CONST_VECTOR.
29134 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
29136 PR middle-end/83575
29137 * predict.c (force_edge_cold): Handle in more sane way edges
29138 with no prediction.
29140 2018-01-09 Carl Love <cel@us.ibm.com>
29142 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
29144 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
29145 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
29146 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
29147 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
29148 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
29149 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
29150 * config/rs6000/rs6000-protos.h: Add extern defition for
29151 rs6000_generate_float2_double_code.
29152 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
29154 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
29155 (float2_v2df): Add define_expand.
29157 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
29160 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
29161 op_mode in the force_to_mode call.
29163 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
29165 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
29166 instead of checking each element individually.
29167 (aarch64_evpc_uzp): Likewise.
29168 (aarch64_evpc_zip): Likewise.
29169 (aarch64_evpc_ext): Likewise.
29170 (aarch64_evpc_rev): Likewise.
29171 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
29172 instead of checking each element individually. Return true without
29174 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
29175 whether all selected elements come from the same input, instead of
29176 checking each element individually. Remove calls to gen_rtx_REG,
29177 start_sequence and end_sequence and instead assert that no rtl is
29180 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
29182 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
29183 order of HIGH and CONST checks.
29185 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
29187 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
29188 if the destination isn't an SSA_NAME.
29190 2018-01-09 Richard Biener <rguenther@suse.de>
29192 PR tree-optimization/83668
29193 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
29195 (canonicalize_loop_form): ... here, renamed from ...
29196 (canonicalize_loop_closed_ssa_form): ... this and amended to
29197 swap successor edges for loop exit blocks to make us use
29198 the RPO order we need for initial schedule generation.
29200 2018-01-09 Joseph Myers <joseph@codesourcery.com>
29202 PR tree-optimization/64811
29203 * match.pd: When optimizing comparisons with Inf, avoid
29204 introducing or losing exceptions from comparisons with NaN.
29206 2018-01-09 Martin Liska <mliska@suse.cz>
29209 * asan.c (shadow_mem_size): Add gcc_assert.
29211 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
29213 Don't save registers in main().
29216 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
29217 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
29218 * config/avr/avr.c (avr_set_current_function): Don't error if
29219 naked, OS_task or OS_main are specified at the same time.
29220 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
29222 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
29224 * common/config/avr/avr-common.c (avr_option_optimization_table):
29225 Switch on -mmain-is-OS_task for optimizing compilations.
29227 2018-01-09 Richard Biener <rguenther@suse.de>
29229 PR tree-optimization/83572
29230 * graphite.c: Include cfganal.h.
29231 (graphite_transform_loops): Connect infinite loops to exit
29232 and remove fake edges at the end.
29234 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
29236 * ipa-inline.c (edge_badness): Revert accidental checkin.
29238 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
29241 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
29242 symbols; not inline clones.
29244 2018-01-09 Jakub Jelinek <jakub@redhat.com>
29247 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
29248 hard registers. Formatting fixes.
29250 PR preprocessor/83722
29251 * gcc.c (try_generate_repro): Pass
29252 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
29253 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
29256 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
29257 Kito Cheng <kito.cheng@gmail.com>
29259 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
29260 (riscv_leaf_function_p): Delete.
29261 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
29263 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
29265 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
29267 (do_ifelse): New function.
29268 (do_isel): New function.
29269 (do_sub3): New function.
29270 (do_add3): New function.
29271 (do_load_mask_compare): New function.
29272 (do_overlap_load_compare): New function.
29273 (expand_compare_loop): New function.
29274 (expand_block_compare): Call expand_compare_loop() when appropriate.
29275 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
29276 option description.
29277 (-mblock-compare-inline-loop-limit): New option.
29279 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
29282 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
29283 Reverse order of second and third operands in first alternative.
29284 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
29285 of first and second elements in UNSPEC_VPERMR vector.
29286 (altivec_expand_vec_perm_le): Likewise.
29288 2018-01-08 Jeff Law <law@redhat.com>
29290 PR rtl-optimizatin/81308
29291 * tree-switch-conversion.c (cfg_altered): New file scoped static.
29292 (process_switch): If group_case_labels makes a change, then set
29294 (pass_convert_switch::execute): If a switch is converted, then
29295 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
29297 PR rtl-optimization/81308
29298 * recog.c (split_all_insns): Conditionally cleanup the CFG after
29301 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
29303 PR target/83663 - Revert r255946
29304 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
29305 generation for cases where splatting a value is not useful.
29306 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
29307 across a vec_duplicate and a paradoxical subreg forming a vector
29308 mode to a vec_concat.
29310 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29312 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
29313 -march=armv8.3-a variants.
29314 * config/arm/t-multilib: Likewise.
29315 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
29317 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
29319 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
29321 (cceq_ior_compare_complement): Give it a name so I can use it, and
29322 change boolean_or_operator predicate to boolean_operator so it can
29323 be used to generate a crand.
29324 (eqne): New code iterator.
29325 (bd/bd_neg): New code_attrs.
29326 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
29327 a single define_insn.
29328 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
29329 decrement (bdnzt/bdnzf/bdzt/bdzf).
29330 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
29331 with the new names of the branch decrement patterns, and added the
29332 names of the branch decrement conditional patterns.
29334 2018-01-08 Richard Biener <rguenther@suse.de>
29336 PR tree-optimization/83563
29337 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
29340 2018-01-08 Richard Biener <rguenther@suse.de>
29342 PR middle-end/83713
29343 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
29345 2018-01-08 Richard Biener <rguenther@suse.de>
29347 PR tree-optimization/83685
29348 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
29349 references to abnormals.
29351 2018-01-08 Richard Biener <rguenther@suse.de>
29354 * dwarf2out.c (output_indirect_strings): Handle empty
29355 skeleton_debug_str_hash.
29356 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
29358 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
29360 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
29361 (emit_store_direct): Likewise.
29362 (arc_trampoline_adjust_address): Likewise.
29363 (arc_asm_trampoline_template): New function.
29364 (arc_initialize_trampoline): Use asm_trampoline_template.
29365 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
29366 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
29367 * config/arc/arc.md (flush_icache): Delete pattern.
29369 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
29371 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
29372 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
29375 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
29378 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
29379 by not USED_FOR_TARGET.
29380 (make_pass_resolve_sw_modes): Likewise.
29382 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
29384 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
29387 2018-01-08 Richard Biener <rguenther@suse.de>
29389 PR middle-end/83580
29390 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
29392 2018-01-08 Richard Biener <rguenther@suse.de>
29394 PR middle-end/83517
29395 * match.pd ((t * 2) / 2) -> t): Add missing :c.
29397 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
29399 PR middle-end/81897
29400 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
29401 basic blocks with a small number of successors.
29402 (convert_control_dep_chain_into_preds): Improve handling of
29404 (dump_predicates): Split apart into...
29405 (dump_pred_chain): ...here...
29406 (dump_pred_info): ...and here.
29407 (can_one_predicate_be_invalidated_p): Add debugging printfs.
29408 (can_chain_union_be_invalidated_p): Improve check for invalidation
29410 (uninit_uses_cannot_happen): Avoid unnecessary if
29411 convert_control_dep_chain_into_preds yielded nothing.
29413 2018-01-06 Martin Sebor <msebor@redhat.com>
29415 PR tree-optimization/83640
29416 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
29417 subtracting negative offset from size.
29418 (builtin_access::overlap): Adjust offset bounds of the access to fall
29419 within the size of the object if possible.
29421 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
29423 PR rtl-optimization/83699
29424 * expmed.c (extract_bit_field_1): Restrict the vector usage of
29425 extract_bit_field_as_subreg to cases in which the extracted
29426 value is also a vector.
29428 * lra-constraints.c (process_alt_operands): Test for the equivalence
29429 substitutions when detecting a possible reload cycle.
29431 2018-01-06 Jakub Jelinek <jakub@redhat.com>
29434 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
29435 by default if flag_selective_schedling{,2}. Formatting fixes.
29437 PR rtl-optimization/83682
29438 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
29439 if it has non-VECTOR_MODE element mode.
29440 (vec_duplicate_p): Likewise.
29442 PR middle-end/83694
29443 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
29444 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
29446 2018-01-05 Jakub Jelinek <jakub@redhat.com>
29449 * config/i386/i386-builtin.def
29450 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
29451 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
29452 Require also OPTION_MASK_ISA_AVX512F in addition to
29453 OPTION_MASK_ISA_GFNI.
29454 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
29455 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
29456 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
29457 to OPTION_MASK_ISA_GFNI.
29458 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
29459 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
29460 OPTION_MASK_ISA_AVX512BW.
29461 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
29462 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
29463 addition to OPTION_MASK_ISA_GFNI.
29464 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
29465 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
29466 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
29467 to OPTION_MASK_ISA_GFNI.
29468 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
29469 a requirement for all ISAs rather than any of them with a few
29471 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
29473 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
29474 bitmasks to be enabled with 3 exceptions, instead of requiring any
29475 enabled ISA with lots of exceptions.
29476 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
29477 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
29478 Change avx512bw in isa attribute to avx512f.
29479 * config/i386/sgxintrin.h: Add license boilerplate.
29480 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
29481 to __AVX512F__ and __AVX512VL to __AVX512VL__.
29482 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
29483 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
29485 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
29486 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
29487 temporarily sse2 rather than sse if not enabled already.
29490 * config/i386/sse.md (VI248_VLBW): Rename to ...
29491 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
29492 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
29493 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
29494 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
29495 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
29496 mode iterator instead of VI248_VLBW.
29498 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
29500 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
29501 (record_modified): Skip clobbers; add debug output.
29502 (param_change_prob): Use sreal frequencies.
29504 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
29506 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
29507 punt for user-aligned variables.
29509 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
29511 * tree-chrec.c (chrec_contains_symbols): Return true for
29514 2018-01-05 Sudakshina Das <sudi.das@arm.com>
29517 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
29518 of (x|y) == x for BICS pattern.
29520 2018-01-05 Jakub Jelinek <jakub@redhat.com>
29522 PR tree-optimization/83605
29523 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
29524 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
29527 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
29529 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
29530 * config/epiphany/rtems.h: New file.
29532 2018-01-04 Jakub Jelinek <jakub@redhat.com>
29533 Uros Bizjak <ubizjak@gmail.com>
29536 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
29537 QIreg_operand instead of register_operand predicate.
29538 * config/i386/i386.c (ix86_rop_should_change_byte_p,
29539 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
29540 comments instead of -fmitigate[-_]rop.
29542 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
29545 * cgraphunit.c (symbol_table::compile): Switch to text_section
29546 before calling assembly_start debug hook.
29547 * run-rtl-passes.c (run_rtl_passes): Likewise.
29550 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
29552 * tree-vrp.c (extract_range_from_binary_expr_1): Check
29553 range_int_cst_p rather than !symbolic_range_p before calling
29554 extract_range_from_multiplicative_op_1.
29556 2018-01-04 Jeff Law <law@redhat.com>
29558 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
29559 redundant test in assertion.
29561 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
29563 * doc/rtl.texi: Document machine_mode wrapper classes.
29565 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
29567 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
29568 using tree_to_uhwi.
29570 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
29572 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
29573 the VEC_PERM_EXPR fold to fail.
29575 2018-01-04 Jakub Jelinek <jakub@redhat.com>
29578 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
29579 to switched_sections.
29581 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
29584 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
29585 test for d.testing.
29587 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
29590 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
29591 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
29593 2018-01-04 Jakub Jelinek <jakub@redhat.com>
29596 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
29597 is BLKmode and bitpos not zero or mode change is needed.
29599 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
29602 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
29605 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
29608 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
29609 instead of MULT rtx. Update all corresponding splitters.
29610 (*saddl_se): Ditto.
29611 (*ssub<modesuffix>): Ditto.
29612 (*ssubl_se): Ditto.
29613 (*cmp_sadd_di): Update split patterns.
29614 (*cmp_sadd_si): Ditto.
29615 (*cmp_sadd_sidi): Ditto.
29616 (*cmp_ssub_di): Ditto.
29617 (*cmp_ssub_si): Ditto.
29618 (*cmp_ssub_sidi): Ditto.
29619 * config/alpha/predicates.md (const23_operand): New predicate.
29620 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
29621 Look for ASHIFT, not MULT inner operand.
29622 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
29624 2018-01-04 Martin Liska <mliska@suse.cz>
29626 PR gcov-profile/83669
29627 * gcov.c (output_intermediate_file): Add version to intermediate
29629 * doc/gcov.texi: Document new field 'version' in intermediate
29630 file format. Fix location of '-k' option of gcov command.
29632 2018-01-04 Martin Liska <mliska@suse.cz>
29635 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
29637 2018-01-04 Jakub Jelinek <jakub@redhat.com>
29639 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
29641 2018-01-03 Martin Sebor <msebor@redhat.com>
29643 PR tree-optimization/83655
29644 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
29645 checking calls with invalid arguments.
29647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29649 * tree-vect-stmts.c (vect_get_store_rhs): New function.
29650 (vectorizable_mask_load_store): Delete.
29651 (vectorizable_call): Return false for masked loads and stores.
29652 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
29653 instead of gimple_assign_rhs1.
29654 (vectorizable_load): Handle IFN_MASK_LOAD.
29655 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
29657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29659 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
29661 (vectorizable_mask_load_store): ...here.
29662 (vectorizable_load): ...and here.
29664 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29666 * tree-vect-stmts.c (vect_build_all_ones_mask)
29667 (vect_build_zero_merge_argument): New functions, split out from...
29668 (vectorizable_load): ...here.
29670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29672 * tree-vect-stmts.c (vect_check_store_rhs): New function,
29674 (vectorizable_mask_load_store): ...here.
29675 (vectorizable_store): ...and here.
29677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29679 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
29681 (vectorizable_mask_load_store): ...here.
29683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29685 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
29686 (vect_model_store_cost): Take a vec_load_store_type instead of a
29688 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
29689 (vect_model_store_cost): Take a vec_load_store_type instead of a
29691 (vectorizable_mask_load_store): Update accordingly.
29692 (vectorizable_store): Likewise.
29693 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
29695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29697 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
29698 IFN_MASK_LOAD calls here rather than...
29699 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
29701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29702 Alan Hayward <alan.hayward@arm.com>
29703 David Sherwood <david.sherwood@arm.com>
29705 * expmed.c (extract_bit_field_1): For vector extracts,
29706 fall back to extract_bit_field_as_subreg if vec_extract
29709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29710 Alan Hayward <alan.hayward@arm.com>
29711 David Sherwood <david.sherwood@arm.com>
29713 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
29714 they are variable or constant sized.
29715 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
29716 slots for constant-sized data.
29718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29719 Alan Hayward <alan.hayward@arm.com>
29720 David Sherwood <david.sherwood@arm.com>
29722 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
29723 handling COND_EXPRs with boolean comparisons, try to find a better
29724 basis for the mask type than the boolean itself.
29726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29728 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
29729 is calculated and how it can be overridden.
29730 * genmodes.c (max_bitsize_mode_any_mode): New variable.
29731 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
29733 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
29736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29737 Alan Hayward <alan.hayward@arm.com>
29738 David Sherwood <david.sherwood@arm.com>
29740 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
29741 Remove the mode argument.
29742 (aarch64_simd_valid_immediate): Remove the mode and inverse
29744 * config/aarch64/iterators.md (bitsize): New iterator.
29745 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
29746 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
29747 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
29748 aarch64_simd_valid_immediate.
29749 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
29750 (aarch64_reg_or_bic_imm): Likewise.
29751 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
29752 with an insn_type enum and msl with a modifier_type enum.
29753 Replace element_width with a scalar_mode. Change the shift
29754 to unsigned int. Add constructors for scalar_float_mode and
29755 scalar_int_mode elements.
29756 (aarch64_vect_float_const_representable_p): Delete.
29757 (aarch64_can_const_movi_rtx_p)
29758 (aarch64_simd_scalar_immediate_valid_for_move)
29759 (aarch64_simd_make_constant): Update call to
29760 aarch64_simd_valid_immediate.
29761 (aarch64_advsimd_valid_immediate_hs): New function.
29762 (aarch64_advsimd_valid_immediate): Likewise.
29763 (aarch64_simd_valid_immediate): Remove mode and inverse
29764 arguments. Rewrite to use the above. Use const_vec_duplicate_p
29765 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
29766 and aarch64_float_const_representable_p on the result.
29767 (aarch64_output_simd_mov_immediate): Remove mode argument.
29768 Update call to aarch64_simd_valid_immediate and use of
29769 simd_immediate_info.
29770 (aarch64_output_scalar_simd_mov_immediate): Update call
29773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29774 Alan Hayward <alan.hayward@arm.com>
29775 David Sherwood <david.sherwood@arm.com>
29777 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
29778 (mode_nunits): Likewise CONST_MODE_NUNITS.
29779 * machmode.def (ADJUST_NUNITS): Document.
29780 * genmodes.c (mode_data::need_nunits_adj): New field.
29781 (blank_mode): Update accordingly.
29782 (adj_nunits): New variable.
29783 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
29785 (emit_mode_size_inline): Set need_bytesize_adj for all modes
29786 listed in adj_nunits.
29787 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
29788 listed in adj_nunits. Don't emit case statements for such modes.
29789 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
29790 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
29791 nothing if adj_nunits is nonnull.
29792 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
29793 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
29794 (emit_mode_fbit): Update use of print_maybe_const_decl.
29795 (emit_move_size): Likewise. Treat the array as non-const
29797 (emit_mode_adjustments): Handle adj_nunits.
29799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29801 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
29802 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
29803 (VECTOR_MODES): Use it.
29804 (make_vector_modes): Take the prefix as an argument.
29806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29807 Alan Hayward <alan.hayward@arm.com>
29808 David Sherwood <david.sherwood@arm.com>
29810 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
29811 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
29812 for MODE_VECTOR_BOOL.
29813 * machmode.def (VECTOR_BOOL_MODE): Document.
29814 * genmodes.c (VECTOR_BOOL_MODE): New macro.
29815 (make_vector_bool_mode): New function.
29816 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
29818 * lto-streamer-in.c (lto_input_mode_table): Likewise.
29819 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
29821 * stor-layout.c (int_mode_for_mode): Likewise.
29822 * tree.c (build_vector_type_for_mode): Likewise.
29823 * varasm.c (output_constant_pool_2): Likewise.
29824 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
29825 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
29826 for MODE_VECTOR_BOOL.
29827 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
29828 of mode class checks.
29829 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
29830 instead of a list of mode class checks.
29831 (expand_vector_scalar_condition): Likewise.
29832 (type_for_widest_vector_mode): Handle BImode as an inner mode.
29834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29835 Alan Hayward <alan.hayward@arm.com>
29836 David Sherwood <david.sherwood@arm.com>
29838 * machmode.h (mode_size): Change from unsigned short to
29840 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
29841 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
29842 or if measurement_type is not polynomial.
29843 (fixed_size_mode::includes_p): Check for constant-sized modes.
29844 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
29845 return a poly_uint16 rather than an unsigned short.
29846 (emit_mode_size): Change the type of mode_size from unsigned short
29847 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
29848 (emit_mode_adjustments): Cope with polynomial vector sizes.
29849 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
29851 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
29853 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
29854 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
29855 * caller-save.c (setup_save_areas): Likewise.
29856 (replace_reg_with_saved_mem): Likewise.
29857 * calls.c (emit_library_call_value_1): Likewise.
29858 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
29859 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
29860 (gen_lowpart_for_combine): Likewise.
29861 * convert.c (convert_to_integer_1): Likewise.
29862 * cse.c (equiv_constant, cse_insn): Likewise.
29863 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
29864 (cselib_subst_to_values): Likewise.
29865 * dce.c (word_dce_process_block): Likewise.
29866 * df-problems.c (df_word_lr_mark_ref): Likewise.
29867 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
29868 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
29869 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
29870 (rtl_for_decl_location): Likewise.
29871 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
29872 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
29873 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
29874 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
29875 (expand_expr_real_1): Likewise.
29876 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
29877 (pad_below): Likewise.
29878 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
29879 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
29880 * ira.c (get_subreg_tracking_sizes): Likewise.
29881 * ira-build.c (ira_create_allocno_objects): Likewise.
29882 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
29883 (ira_sort_regnos_for_alter_reg): Likewise.
29884 * ira-costs.c (record_operand_costs): Likewise.
29885 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
29886 (resolve_simple_move): Likewise.
29887 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
29888 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
29889 (lra_constraints): Likewise.
29890 (CONST_POOL_OK_P): Reject variable-sized modes.
29891 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
29892 (add_pseudo_to_slot, lra_spill): Likewise.
29893 * omp-low.c (omp_clause_aligned_alignment): Likewise.
29894 * optabs-query.c (get_best_extraction_insn): Likewise.
29895 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
29896 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
29897 (expand_mult_highpart, valid_multiword_target_p): Likewise.
29898 * recog.c (offsettable_address_addr_space_p): Likewise.
29899 * regcprop.c (maybe_mode_change): Likewise.
29900 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
29901 * regrename.c (build_def_use): Likewise.
29902 * regstat.c (dump_reg_info): Likewise.
29903 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
29904 (find_reloads, find_reloads_subreg_address): Likewise.
29905 * reload1.c (eliminate_regs_1): Likewise.
29906 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
29907 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
29908 (simplify_binary_operation_1, simplify_subreg): Likewise.
29909 * targhooks.c (default_function_arg_padding): Likewise.
29910 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
29911 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
29912 (verify_gimple_assign_ternary): Likewise.
29913 * tree-inline.c (estimate_move_cost): Likewise.
29914 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29915 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
29916 (get_address_cost_ainc): Likewise.
29917 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
29918 (vect_supportable_dr_alignment): Likewise.
29919 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
29920 (vectorizable_reduction): Likewise.
29921 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
29922 (vectorizable_operation, vectorizable_load): Likewise.
29923 * tree.c (build_same_sized_truth_vector_type): Likewise.
29924 * valtrack.c (cleanup_auto_inc_dec): Likewise.
29925 * var-tracking.c (emit_note_insn_var_location): Likewise.
29926 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
29927 (ADDR_VEC_ALIGN): Likewise.
29929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29930 Alan Hayward <alan.hayward@arm.com>
29931 David Sherwood <david.sherwood@arm.com>
29933 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
29935 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
29936 or if measurement_type is polynomial.
29937 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
29938 * combine.c (make_extraction): Likewise.
29939 * dse.c (find_shift_sequence): Likewise.
29940 * dwarf2out.c (mem_loc_descriptor): Likewise.
29941 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
29942 (extract_bit_field, extract_low_bits): Likewise.
29943 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
29944 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
29945 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
29946 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
29947 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
29948 * reload.c (find_reloads): Likewise.
29949 * reload1.c (alter_reg): Likewise.
29950 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
29951 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
29952 * tree-if-conv.c (predicate_mem_writes): Likewise.
29953 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
29954 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
29955 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
29956 * valtrack.c (dead_debug_insert_temp): Likewise.
29957 * varasm.c (mergeable_constant_section): Likewise.
29958 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
29960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29961 Alan Hayward <alan.hayward@arm.com>
29962 David Sherwood <david.sherwood@arm.com>
29964 * expr.c (expand_assignment): Cope with polynomial mode sizes
29965 when assigning to a CONCAT.
29967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
29968 Alan Hayward <alan.hayward@arm.com>
29969 David Sherwood <david.sherwood@arm.com>
29971 * machmode.h (mode_precision): Change from unsigned short to
29973 (mode_to_precision): Return a poly_uint16 rather than an unsigned
29975 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
29976 or if measurement_type is not polynomial.
29977 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
29978 in which the mode is already known to be a scalar_int_mode.
29979 * genmodes.c (emit_mode_precision): Change the type of mode_precision
29980 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
29982 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
29983 for GET_MODE_PRECISION.
29984 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
29985 for GET_MODE_PRECISION.
29986 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
29988 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
29989 (expand_field_assignment, make_extraction): Likewise.
29990 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
29991 (get_last_value): Likewise.
29992 * convert.c (convert_to_integer_1): Likewise.
29993 * cse.c (cse_insn): Likewise.
29994 * expr.c (expand_expr_real_1): Likewise.
29995 * lra-constraints.c (simplify_operand_subreg): Likewise.
29996 * optabs-query.c (can_atomic_load_p): Likewise.
29997 * optabs.c (expand_atomic_load): Likewise.
29998 (expand_atomic_store): Likewise.
29999 * ree.c (combine_reaching_defs): Likewise.
30000 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
30001 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
30002 * tree.h (type_has_mode_precision_p): Likewise.
30003 * ubsan.c (instrument_si_overflow): Likewise.
30005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30006 Alan Hayward <alan.hayward@arm.com>
30007 David Sherwood <david.sherwood@arm.com>
30009 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
30010 polynomial numbers of units.
30011 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
30012 (valid_vector_subparts_p): New function.
30013 (build_vector_type): Remove temporary shim and take the number
30014 of units as a poly_uint64 rather than an int.
30015 (build_opaque_vector_type): Take the number of units as a
30016 poly_uint64 rather than an int.
30017 * tree.c (build_vector_from_ctor): Handle polynomial
30018 TYPE_VECTOR_SUBPARTS.
30019 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
30020 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
30021 (build_vector_from_val): If the number of units is variable,
30022 use build_vec_duplicate_cst for constant operands and
30023 VEC_DUPLICATE_EXPR otherwise.
30024 (make_vector_type): Remove temporary is_constant ().
30025 (build_vector_type, build_opaque_vector_type): Take the number of
30026 units as a poly_uint64 rather than an int.
30027 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
30029 * cfgexpand.c (expand_debug_expr): Likewise.
30030 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
30031 (store_constructor, expand_expr_real_1): Likewise.
30032 (const_scalar_mask_from_tree): Likewise.
30033 * fold-const-call.c (fold_const_reduction): Likewise.
30034 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
30035 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
30036 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
30037 (fold_relational_const): Likewise.
30038 (native_interpret_vector): Likewise. Change the size from an
30039 int to an unsigned int.
30040 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
30041 TYPE_VECTOR_SUBPARTS.
30042 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
30043 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
30044 duplicating a non-constant operand into a variable-length vector.
30045 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
30046 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
30047 * ipa-icf.c (sem_variable::equals): Likewise.
30048 * match.pd: Likewise.
30049 * omp-simd-clone.c (simd_clone_subparts): Likewise.
30050 * print-tree.c (print_node): Likewise.
30051 * stor-layout.c (layout_type): Likewise.
30052 * targhooks.c (default_builtin_vectorization_cost): Likewise.
30053 * tree-cfg.c (verify_gimple_comparison): Likewise.
30054 (verify_gimple_assign_binary): Likewise.
30055 (verify_gimple_assign_ternary): Likewise.
30056 (verify_gimple_assign_single): Likewise.
30057 * tree-pretty-print.c (dump_generic_node): Likewise.
30058 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30059 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
30060 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
30061 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
30062 (vect_shift_permute_load_chain): Likewise.
30063 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
30064 (expand_vector_condition, optimize_vector_constructor): Likewise.
30065 (lower_vec_perm, get_compute_type): Likewise.
30066 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
30067 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
30068 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
30069 (vect_recog_mask_conversion_pattern): Likewise.
30070 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
30071 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
30072 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30073 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
30074 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
30075 (vectorizable_shift, vectorizable_operation, vectorizable_store)
30076 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
30077 (supportable_widening_operation): Likewise.
30078 (supportable_narrowing_operation): Likewise.
30079 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
30081 * varasm.c (output_constant): Likewise.
30083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30084 Alan Hayward <alan.hayward@arm.com>
30085 David Sherwood <david.sherwood@arm.com>
30087 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
30088 so that both the length == 3 and length != 3 cases set up their
30089 own permute vectors. Add comments explaining why we know the
30090 number of elements is constant.
30091 (vect_permute_load_chain): Likewise.
30093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30094 Alan Hayward <alan.hayward@arm.com>
30095 David Sherwood <david.sherwood@arm.com>
30097 * machmode.h (mode_nunits): Change from unsigned char to
30099 (ONLY_FIXED_SIZE_MODES): New macro.
30100 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
30101 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
30102 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
30104 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
30105 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
30106 or if measurement_type is not polynomial.
30107 * genmodes.c (ZERO_COEFFS): New macro.
30108 (emit_mode_nunits_inline): Make mode_nunits_inline return a
30110 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
30111 Use ZERO_COEFFS when emitting initializers.
30112 * data-streamer.h (bp_pack_poly_value): New function.
30113 (bp_unpack_poly_value): Likewise.
30114 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30115 for GET_MODE_NUNITS.
30116 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30117 for GET_MODE_NUNITS.
30118 * tree.c (make_vector_type): Remove temporary shim and make
30119 the real function take the number of units as a poly_uint64
30120 rather than an int.
30121 (build_vector_type_for_mode): Handle polynomial nunits.
30122 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
30123 * emit-rtl.c (const_vec_series_p_1): Likewise.
30124 (gen_rtx_CONST_VECTOR): Likewise.
30125 * fold-const.c (test_vec_duplicate_folding): Likewise.
30126 * genrecog.c (validate_pattern): Likewise.
30127 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
30128 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
30129 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
30130 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
30131 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
30132 * rtlanal.c (subreg_get_info): Likewise.
30133 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
30134 (vect_grouped_load_supported): Likewise.
30135 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
30136 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
30137 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
30138 (simplify_const_unary_operation, simplify_binary_operation_1)
30139 (simplify_const_binary_operation, simplify_ternary_operation)
30140 (test_vector_ops_duplicate, test_vector_ops): Likewise.
30141 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
30142 instead of CONST_VECTOR_NUNITS.
30143 * varasm.c (output_constant_pool_2): Likewise.
30144 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
30145 explicit-encoded elements in the XVEC for variable-length vectors.
30147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30149 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
30151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30152 Alan Hayward <alan.hayward@arm.com>
30153 David Sherwood <david.sherwood@arm.com>
30155 * coretypes.h (fixed_size_mode): Declare.
30156 (fixed_size_mode_pod): New typedef.
30157 * builtins.h (target_builtins::x_apply_args_mode)
30158 (target_builtins::x_apply_result_mode): Change type to
30159 fixed_size_mode_pod.
30160 * builtins.c (apply_args_size, apply_result_size, result_vector)
30161 (expand_builtin_apply_args_1, expand_builtin_apply)
30162 (expand_builtin_return): Update accordingly.
30164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30166 * cse.c (hash_rtx_cb): Hash only the encoded elements.
30167 * cselib.c (cselib_hash_rtx): Likewise.
30168 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
30169 CONST_VECTOR encoding.
30171 2018-01-03 Jakub Jelinek <jakub@redhat.com>
30172 Jeff Law <law@redhat.com>
30175 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
30176 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
30177 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
30178 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
30181 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
30182 explicitly probe *sp in a noreturn function if there were any callee
30183 register saves or frame pointer is needed.
30185 2018-01-03 Jakub Jelinek <jakub@redhat.com>
30188 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
30189 BLKmode for ternary, binary or unary expressions.
30192 * var-tracking.c (delete_vta_debug_insn): New inline function.
30193 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
30194 insns from get_insns () to NULL instead of each bb separately.
30195 Use delete_vta_debug_insn. No longer static.
30196 (vt_debug_insns_local, variable_tracking_main_1): Adjust
30197 delete_vta_debug_insns callers.
30198 * rtl.h (delete_vta_debug_insns): Declare.
30199 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
30200 instead of variable_tracking_main.
30202 2018-01-03 Martin Sebor <msebor@redhat.com>
30204 PR tree-optimization/83603
30205 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
30206 arguments past the endof the argument list in functions declared
30207 without a prototype.
30208 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
30209 Avoid checking when arguments are null.
30211 2018-01-03 Martin Sebor <msebor@redhat.com>
30214 * doc/extend.texi (attribute const): Fix a typo.
30215 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
30216 issuing -Wsuggest-attribute for void functions.
30218 2018-01-03 Martin Sebor <msebor@redhat.com>
30220 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
30221 offset_int::from instead of wide_int::to_shwi.
30222 (maybe_diag_overlap): Remove assertion.
30223 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
30224 * gimple-ssa-sprintf.c (format_directive): Same.
30225 (parse_directive): Same.
30226 (sprintf_dom_walker::compute_format_length): Same.
30227 (try_substitute_return_value): Same.
30229 2018-01-03 Jeff Law <law@redhat.com>
30231 PR middle-end/83654
30232 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
30233 non-constant residual for zero at runtime and avoid probing in
30234 that case. Reorganize code for trailing problem to mirror handling
30237 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
30239 PR tree-optimization/83501
30240 * tree-ssa-strlen.c (get_string_cst): New.
30241 (handle_char_store): Call get_string_cst.
30243 2018-01-03 Martin Liska <mliska@suse.cz>
30245 PR tree-optimization/83593
30246 * tree-ssa-strlen.c: Include tree-cfg.h.
30247 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
30248 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
30249 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
30251 (strlen_dom_walker::before_dom_children): Call
30252 gimple_purge_dead_eh_edges. Dump tranformation with details
30254 (strlen_dom_walker::before_dom_children): Update call by adding
30255 new argument cleanup_eh.
30256 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
30258 2018-01-03 Martin Liska <mliska@suse.cz>
30261 * cif-code.def (VARIADIC_THUNK): New enum value.
30262 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
30265 2018-01-03 Jan Beulich <jbeulich@suse.com>
30267 * sse.md (mov<mode>_internal): Tighten condition for when to use
30268 vmovdqu<ssescalarsize> for TI and OI modes.
30270 2018-01-03 Jakub Jelinek <jakub@redhat.com>
30272 Update copyright years.
30274 2018-01-03 Martin Liska <mliska@suse.cz>
30277 * ipa-visibility.c (function_and_variable_visibility): Skip
30278 functions with noipa attribure.
30280 2018-01-03 Jakub Jelinek <jakub@redhat.com>
30282 * gcc.c (process_command): Update copyright notice dates.
30283 * gcov-dump.c (print_version): Ditto.
30284 * gcov.c (print_version): Ditto.
30285 * gcov-tool.c (print_version): Ditto.
30286 * gengtype.c (create_file): Ditto.
30287 * doc/cpp.texi: Bump @copying's copyright year.
30288 * doc/cppinternals.texi: Ditto.
30289 * doc/gcc.texi: Ditto.
30290 * doc/gccint.texi: Ditto.
30291 * doc/gcov.texi: Ditto.
30292 * doc/install.texi: Ditto.
30293 * doc/invoke.texi: Ditto.
30295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30297 * vector-builder.h (vector_builder::m_full_nelts): Change from
30298 unsigned int to poly_uint64.
30299 (vector_builder::full_nelts): Update prototype accordingly.
30300 (vector_builder::new_vector): Likewise.
30301 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
30302 (vector_builder::operator ==): Likewise.
30303 (vector_builder::finalize): Likewise.
30304 * int-vector-builder.h (int_vector_builder::int_vector_builder):
30305 Take the number of elements as a poly_uint64 rather than an
30307 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
30308 from unsigned int to poly_uint64.
30309 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
30310 (vec_perm_indices::new_vector): Likewise.
30311 (vec_perm_indices::length): Likewise.
30312 (vec_perm_indices::nelts_per_input): Likewise.
30313 (vec_perm_indices::input_nelts): Likewise.
30314 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
30315 number of elements per input as a poly_uint64 rather than an
30316 unsigned int. Use the original encoding for variable-length
30317 vectors, rather than clamping each individual element.
30318 For the second and subsequent elements in each pattern,
30319 clamp the step and base before clamping their sum.
30320 (vec_perm_indices::series_p): Handle polynomial element counts.
30321 (vec_perm_indices::all_in_range_p): Likewise.
30322 (vec_perm_indices_to_tree): Likewise.
30323 (vec_perm_indices_to_rtx): Likewise.
30324 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
30325 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
30326 (tree_vector_builder::new_binary_operation): Handle polynomial
30327 element counts. Return false if we need to know the number
30328 of elements at compile time.
30329 * fold-const.c (fold_vec_perm): Punt if the number of elements
30330 isn't known at compile time.
30332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30334 * vec-perm-indices.h (vec_perm_builder): Change element type
30335 from HOST_WIDE_INT to poly_int64.
30336 (vec_perm_indices::element_type): Update accordingly.
30337 (vec_perm_indices::clamp): Handle polynomial element_types.
30338 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
30339 (vec_perm_indices::all_in_range_p): Likewise.
30340 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
30342 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
30343 polynomial vec_perm_indices element types.
30344 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
30345 * fold-const.c (fold_vec_perm): Likewise.
30346 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
30347 * tree-vect-generic.c (lower_vec_perm): Likewise.
30348 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
30349 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
30350 element type to HOST_WIDE_INT.
30352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30353 Alan Hayward <alan.hayward@arm.com>
30354 David Sherwood <david.sherwood@arm.com>
30356 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
30357 rather than an int. Use plus_constant.
30358 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
30359 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
30361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30362 Alan Hayward <alan.hayward@arm.com>
30363 David Sherwood <david.sherwood@arm.com>
30365 * calls.c (emit_call_1, expand_call): Change struct_value_size from
30366 a HOST_WIDE_INT to a poly_int64.
30368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30369 Alan Hayward <alan.hayward@arm.com>
30370 David Sherwood <david.sherwood@arm.com>
30372 * calls.c (load_register_parameters): Cope with polynomial
30373 mode sizes. Require a constant size for BLKmode parameters
30374 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
30375 forces a parameter to be padded at the lsb end in order to
30376 fill a complete number of words, require the parameter size
30377 to be ordered wrt UNITS_PER_WORD.
30379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30380 Alan Hayward <alan.hayward@arm.com>
30381 David Sherwood <david.sherwood@arm.com>
30383 * reload1.c (spill_stack_slot_width): Change element type
30384 from unsigned int to poly_uint64_pod.
30385 (alter_reg): Treat mode sizes as polynomial.
30387 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30388 Alan Hayward <alan.hayward@arm.com>
30389 David Sherwood <david.sherwood@arm.com>
30391 * reload.c (complex_word_subreg_p): New function.
30392 (reload_inner_reg_of_subreg, push_reload): Use it.
30394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30395 Alan Hayward <alan.hayward@arm.com>
30396 David Sherwood <david.sherwood@arm.com>
30398 * lra-constraints.c (process_alt_operands): Reject matched
30399 operands whose sizes aren't ordered.
30400 (match_reload): Refer to this check here.
30402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30403 Alan Hayward <alan.hayward@arm.com>
30404 David Sherwood <david.sherwood@arm.com>
30406 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
30407 that the mode size is in the set {1, 2, 4, 8, 16}.
30409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30410 Alan Hayward <alan.hayward@arm.com>
30411 David Sherwood <david.sherwood@arm.com>
30413 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
30414 Use plus_constant instead of gen_rtx_PLUS.
30416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30417 Alan Hayward <alan.hayward@arm.com>
30418 David Sherwood <david.sherwood@arm.com>
30420 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
30421 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
30422 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
30423 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
30424 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
30425 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
30426 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
30427 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
30428 * config/i386/i386.c (ix86_push_rounding): ...this new function.
30429 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
30431 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
30432 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
30433 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
30434 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
30435 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
30436 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
30437 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
30438 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
30439 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
30440 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
30442 * expr.c (emit_move_resolve_push): Treat the input and result
30443 of PUSH_ROUNDING as a poly_int64.
30444 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
30445 (emit_push_insn): Likewise.
30446 * lra-eliminations.c (mark_not_eliminable): Likewise.
30447 * recog.c (push_operand): Likewise.
30448 * reload1.c (elimination_effects): Likewise.
30449 * rtlanal.c (nonzero_bits1): Likewise.
30450 * calls.c (store_one_arg): Likewise. Require the padding to be
30451 known at compile time.
30453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30454 Alan Hayward <alan.hayward@arm.com>
30455 David Sherwood <david.sherwood@arm.com>
30457 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
30458 Use plus_constant instead of gen_rtx_PLUS.
30460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30461 Alan Hayward <alan.hayward@arm.com>
30462 David Sherwood <david.sherwood@arm.com>
30464 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
30465 rather than an int.
30467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30468 Alan Hayward <alan.hayward@arm.com>
30469 David Sherwood <david.sherwood@arm.com>
30471 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
30472 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
30473 via stack temporaries. Treat the mode size as polynomial too.
30475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30476 Alan Hayward <alan.hayward@arm.com>
30477 David Sherwood <david.sherwood@arm.com>
30479 * expr.c (expand_expr_real_2): When handling conversions involving
30480 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
30481 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
30482 as a poly_uint64 too.
30484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30485 Alan Hayward <alan.hayward@arm.com>
30486 David Sherwood <david.sherwood@arm.com>
30488 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
30490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30491 Alan Hayward <alan.hayward@arm.com>
30492 David Sherwood <david.sherwood@arm.com>
30494 * combine.c (can_change_dest_mode): Handle polynomial
30495 REGMODE_NATURAL_SIZE.
30496 * expmed.c (store_bit_field_1): Likewise.
30497 * expr.c (store_constructor): Likewise.
30498 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
30499 and polynomial REGMODE_NATURAL_SIZE.
30500 (gen_lowpart_common): Likewise.
30501 * reginfo.c (record_subregs_of_mode): Likewise.
30502 * rtlanal.c (read_modify_subreg_p): Likewise.
30504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30505 Alan Hayward <alan.hayward@arm.com>
30506 David Sherwood <david.sherwood@arm.com>
30508 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
30509 numbers of elements.
30511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30512 Alan Hayward <alan.hayward@arm.com>
30513 David Sherwood <david.sherwood@arm.com>
30515 * match.pd: Cope with polynomial numbers of vector elements.
30517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30518 Alan Hayward <alan.hayward@arm.com>
30519 David Sherwood <david.sherwood@arm.com>
30521 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
30522 in a POINTER_PLUS_EXPR.
30524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30525 Alan Hayward <alan.hayward@arm.com>
30526 David Sherwood <david.sherwood@arm.com>
30528 * omp-simd-clone.c (simd_clone_subparts): New function.
30529 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
30530 (ipa_simd_modify_function_body): Likewise.
30532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30533 Alan Hayward <alan.hayward@arm.com>
30534 David Sherwood <david.sherwood@arm.com>
30536 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
30537 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
30538 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
30539 (expand_vector_condition, vector_element): Likewise.
30540 (subparts_gt): New function.
30541 (get_compute_type): Use subparts_gt.
30542 (count_type_subparts): Delete.
30543 (expand_vector_operations_1): Use subparts_gt instead of
30544 count_type_subparts.
30546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30547 Alan Hayward <alan.hayward@arm.com>
30548 David Sherwood <david.sherwood@arm.com>
30550 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
30551 (vect_compile_time_alias): ...this new function. Do the calculation
30552 on poly_ints rather than trees.
30553 (vect_prune_runtime_alias_test_list): Update call accordingly.
30555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30556 Alan Hayward <alan.hayward@arm.com>
30557 David Sherwood <david.sherwood@arm.com>
30559 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
30561 (vect_schedule_slp_instance): Likewise.
30563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30564 Alan Hayward <alan.hayward@arm.com>
30565 David Sherwood <david.sherwood@arm.com>
30567 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
30568 constant and extern definitions for variable-length vectors.
30569 (vect_get_constant_vectors): Note that the number of units
30570 is known to be constant.
30572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30573 Alan Hayward <alan.hayward@arm.com>
30574 David Sherwood <david.sherwood@arm.com>
30576 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
30577 of units as polynomial. Choose between WIDE and NARROW based
30580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30581 Alan Hayward <alan.hayward@arm.com>
30582 David Sherwood <david.sherwood@arm.com>
30584 * tree-vect-stmts.c (simd_clone_subparts): New function.
30585 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
30587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30588 Alan Hayward <alan.hayward@arm.com>
30589 David Sherwood <david.sherwood@arm.com>
30591 * tree-vect-stmts.c (vectorizable_call): Treat the number of
30592 vectors as polynomial. Use build_index_vector for
30593 IFN_GOMP_SIMD_LANE.
30595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30596 Alan Hayward <alan.hayward@arm.com>
30597 David Sherwood <david.sherwood@arm.com>
30599 * tree-vect-stmts.c (get_load_store_type): Treat the number of
30600 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
30601 for variable-length vectors.
30602 (vectorizable_mask_load_store): Treat the number of units as
30603 polynomial, asserting that it is constant if the condition has
30604 already been enforced.
30605 (vectorizable_store, vectorizable_load): Likewise.
30607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30608 Alan Hayward <alan.hayward@arm.com>
30609 David Sherwood <david.sherwood@arm.com>
30611 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
30612 of units as polynomial. Punt if we can't tell at compile time
30613 which vector contains the final result.
30615 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30616 Alan Hayward <alan.hayward@arm.com>
30617 David Sherwood <david.sherwood@arm.com>
30619 * tree-vect-loop.c (vectorizable_induction): Treat the number
30620 of units as polynomial. Punt on SLP inductions. Use an integer
30621 VEC_SERIES_EXPR for variable-length integer reductions. Use a
30622 cast of such a series for variable-length floating-point
30625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30626 Alan Hayward <alan.hayward@arm.com>
30627 David Sherwood <david.sherwood@arm.com>
30629 * tree.h (build_index_vector): Declare.
30630 * tree.c (build_index_vector): New function.
30631 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
30632 of units as polynomial, forcibly converting it to a constant if
30633 vectorizable_reduction has already enforced the condition.
30634 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
30635 to create a {1,2,3,...} vector.
30636 (vectorizable_reduction): Treat the number of units as polynomial.
30637 Choose vectype_in based on the largest scalar element size rather
30638 than the smallest number of units. Enforce the restrictions
30641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30642 Alan Hayward <alan.hayward@arm.com>
30643 David Sherwood <david.sherwood@arm.com>
30645 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
30646 number of units as polynomial.
30648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30649 Alan Hayward <alan.hayward@arm.com>
30650 David Sherwood <david.sherwood@arm.com>
30652 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
30653 * target.def (autovectorize_vector_sizes): Return the vector sizes
30654 by pointer, using vector_sizes rather than a bitmask.
30655 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
30656 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
30657 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
30659 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
30660 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
30661 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
30662 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
30663 * omp-general.c (omp_max_vf): Likewise.
30664 * omp-low.c (omp_clause_aligned_alignment): Likewise.
30665 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
30666 * tree-vect-loop.c (vect_analyze_loop): Likewise.
30667 * tree-vect-slp.c (vect_slp_bb): Likewise.
30668 * doc/tm.texi: Regenerate.
30669 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
30671 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
30672 the vector size as a poly_uint64 rather than an unsigned int.
30673 (current_vector_size): Change from an unsigned int to a poly_uint64.
30674 (get_vectype_for_scalar_type): Update accordingly.
30675 * tree.h (build_truth_vector_type): Take the size and number of
30676 units as a poly_uint64 rather than an unsigned int.
30677 (build_vector_type): Add a temporary overload that takes
30678 the number of units as a poly_uint64 rather than an unsigned int.
30679 * tree.c (make_vector_type): Likewise.
30680 (build_truth_vector_type): Take the number of units as a poly_uint64
30681 rather than an unsigned int.
30683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30684 Alan Hayward <alan.hayward@arm.com>
30685 David Sherwood <david.sherwood@arm.com>
30687 * target.def (get_mask_mode): Take the number of units and length
30688 as poly_uint64s rather than unsigned ints.
30689 * targhooks.h (default_get_mask_mode): Update accordingly.
30690 * targhooks.c (default_get_mask_mode): Likewise.
30691 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
30692 * doc/tm.texi: Regenerate.
30694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30695 Alan Hayward <alan.hayward@arm.com>
30696 David Sherwood <david.sherwood@arm.com>
30698 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
30699 * omp-general.c (omp_max_vf): Likewise.
30700 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
30701 (expand_omp_simd): Handle polynomial safelen.
30702 * omp-low.c (omplow_simd_context): Add a default constructor.
30703 (omplow_simd_context::max_vf): Change from int to poly_uint64.
30704 (lower_rec_simd_input_clauses): Update accordingly.
30705 (lower_rec_input_clauses): Likewise.
30707 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30708 Alan Hayward <alan.hayward@arm.com>
30709 David Sherwood <david.sherwood@arm.com>
30711 * tree-vectorizer.h (vect_nunits_for_cost): New function.
30712 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
30713 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
30714 (vect_analyze_slp_cost): Likewise.
30715 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
30716 (vect_model_load_cost): Likewise.
30718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30719 Alan Hayward <alan.hayward@arm.com>
30720 David Sherwood <david.sherwood@arm.com>
30722 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
30723 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
30724 from an unsigned int * to a poly_uint64_pod *.
30725 (calculate_unrolling_factor): New function.
30726 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
30728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30729 Alan Hayward <alan.hayward@arm.com>
30730 David Sherwood <david.sherwood@arm.com>
30732 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
30733 from an unsigned int to a poly_uint64.
30734 (_loop_vec_info::slp_unrolling_factor): Likewise.
30735 (_loop_vec_info::vectorization_factor): Change from an int
30737 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
30738 (vect_get_num_vectors): New function.
30739 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
30740 (vect_get_num_copies): Use vect_get_num_vectors.
30741 (vect_analyze_data_ref_dependences): Change max_vf from an int *
30742 to an unsigned int *.
30743 (vect_analyze_data_refs): Change min_vf from an int * to a
30745 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
30746 than an unsigned HOST_WIDE_INT.
30747 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
30748 (vect_analyze_data_ref_dependence): Change max_vf from an int *
30749 to an unsigned int *.
30750 (vect_analyze_data_ref_dependences): Likewise.
30751 (vect_compute_data_ref_alignment): Handle polynomial vf.
30752 (vect_enhance_data_refs_alignment): Likewise.
30753 (vect_prune_runtime_alias_test_list): Likewise.
30754 (vect_shift_permute_load_chain): Likewise.
30755 (vect_supportable_dr_alignment): Likewise.
30756 (dependence_distance_ge_vf): Take the vectorization factor as a
30757 poly_uint64 rather than an unsigned HOST_WIDE_INT.
30758 (vect_analyze_data_refs): Change min_vf from an int * to a
30760 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
30761 vfm1 as a poly_uint64 rather than an int. Make the same change
30762 for the returned bound_scalar.
30763 (vect_gen_vector_loop_niters): Handle polynomial vf.
30764 (vect_do_peeling): Likewise. Update call to
30765 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
30766 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
30768 * tree-vect-loop.c (vect_determine_vectorization_factor)
30769 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
30770 (vect_get_known_peeling_cost): Likewise.
30771 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
30772 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
30773 (vect_transform_loop): Likewise. Use the lowest possible VF when
30774 updating the upper bounds of the loop.
30775 (vect_min_worthwhile_factor): Make static. Return an unsigned int
30776 rather than an int.
30777 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
30778 polynomial unroll factors.
30779 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
30780 (vect_make_slp_decision): Likewise.
30781 (vect_supported_load_permutation_p): Likewise, and polynomial
30783 (vect_analyze_slp_cost): Handle polynomial vf.
30784 (vect_slp_analyze_node_operations): Likewise.
30785 (vect_slp_analyze_bb_1): Likewise.
30786 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
30787 than an unsigned HOST_WIDE_INT.
30788 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
30789 (vectorizable_load): Handle polynomial vf.
30790 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
30792 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
30794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30795 Alan Hayward <alan.hayward@arm.com>
30796 David Sherwood <david.sherwood@arm.com>
30798 * match.pd: Handle bit operations involving three constants
30799 and try to fold one pair.
30801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30803 * tree-vect-loop-manip.c: Include gimple-fold.h.
30804 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
30805 niters_maybe_zero parameters. Handle other cases besides a step of 1.
30806 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
30807 Add a path that uses a step of VF instead of 1, but disable it
30809 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
30810 and niters_no_overflow parameters. Update calls to
30811 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
30812 Create a new SSA name if the latter choses to use a ste other
30813 than zero, and return it via niters_vector_mult_vf_var.
30814 * tree-vect-loop.c (vect_transform_loop): Update calls to
30815 vect_do_peeling, vect_gen_vector_loop_niters and
30816 slpeel_make_loop_iterate_ntimes.
30817 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
30818 (vect_gen_vector_loop_niters): Update declarations after above changes.
30820 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
30822 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
30823 128-bit round to integer instructions.
30824 (ceil<mode>2): Likewise.
30825 (btrunc<mode>2): Likewise.
30826 (round<mode>2): Likewise.
30828 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
30830 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
30831 unaligned VSX load/store on P8/P9.
30832 (expand_block_clear): Allow the use of unaligned VSX
30833 load/store on P8/P9.
30835 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
30837 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
30839 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
30840 swap associated with both a load and a store.
30842 2018-01-02 Andrew Waterman <andrew@sifive.com>
30844 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
30845 * config/riscv/riscv.md (clear_cache): Use it.
30847 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
30849 * web.c: Remove out-of-date comment.
30851 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30853 * expr.c (fixup_args_size_notes): Check that any existing
30854 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
30855 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
30856 (emit_single_push_insn): ...here.
30858 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30860 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
30861 (const_vector_encoded_nelts): New function.
30862 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
30863 (const_vector_int_elt, const_vector_elt): Declare.
30864 * emit-rtl.c (const_vector_int_elt_1): New function.
30865 (const_vector_elt): Likewise.
30866 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
30867 of CONST_VECTOR_ELT.
30869 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30871 * expr.c: Include rtx-vector-builder.h.
30872 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
30873 directly on the tree encoding.
30874 (const_vector_from_tree): Likewise.
30875 * optabs.c: Include rtx-vector-builder.h.
30876 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
30877 sequence of "u" values.
30878 * vec-perm-indices.c: Include rtx-vector-builder.h.
30879 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
30880 directly on the vec_perm_indices encoding.
30882 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30884 * doc/rtl.texi (const_vector): Describe new encoding scheme.
30885 * Makefile.in (OBJS): Add rtx-vector-builder.o.
30886 * rtx-vector-builder.h: New file.
30887 * rtx-vector-builder.c: Likewise.
30888 * rtl.h (rtx_def::u2): Add a const_vector field.
30889 (CONST_VECTOR_NPATTERNS): New macro.
30890 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
30891 (CONST_VECTOR_DUPLICATE_P): Likewise.
30892 (CONST_VECTOR_STEPPED_P): Likewise.
30893 (CONST_VECTOR_ENCODED_ELT): Likewise.
30894 (const_vec_duplicate_p): Check for a duplicated vector encoding.
30895 (unwrap_const_vec_duplicate): Likewise.
30896 (const_vec_series_p): Check for a non-duplicated vector encoding.
30897 Say that the function only returns true for integer vectors.
30898 * emit-rtl.c: Include rtx-vector-builder.h.
30899 (gen_const_vec_duplicate_1): Delete.
30900 (gen_const_vector): Call gen_const_vec_duplicate instead of
30901 gen_const_vec_duplicate_1.
30902 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
30903 (gen_const_vec_duplicate): Use rtx_vector_builder.
30904 (gen_const_vec_series): Likewise.
30905 (gen_rtx_CONST_VECTOR): Likewise.
30906 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
30907 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
30908 Build a new vector rather than modifying a CONST_VECTOR in-place.
30909 (handle_special_swappables): Update call accordingly.
30910 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
30911 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
30912 Build a new vector rather than modifying a CONST_VECTOR in-place.
30913 (handle_special_swappables): Update call accordingly.
30915 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30917 * simplify-rtx.c (simplify_const_binary_operation): Use
30918 CONST_VECTOR_ELT instead of XVECEXP.
30920 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30922 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
30923 the selector elements to be different from the data elements
30924 if the selector is a VECTOR_CST.
30925 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
30926 ssizetype for the selector.
30928 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30930 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
30931 before testing each element individually.
30932 * tree-vect-generic.c (lower_vec_perm): Likewise.
30934 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30936 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
30937 * selftest-run-tests.c (selftest::run_tests): Call it.
30938 * vector-builder.h (vector_builder::operator ==): New function.
30939 (vector_builder::operator !=): Likewise.
30940 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
30941 (vec_perm_indices::all_from_input_p): New function.
30942 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
30943 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
30944 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
30945 instead of reading the VECTOR_CST directly. Detect whether both
30946 vector inputs are the same before constructing the vec_perm_indices,
30947 and update the number of inputs argument accordingly. Use the
30948 utility functions added above. Only construct sel2 if we need to.
30950 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30952 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
30953 the broadcast of the low byte.
30954 (expand_mult_highpart): Use an explicit encoding for the permutes.
30955 * optabs-query.c (can_mult_highpart_p): Likewise.
30956 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
30957 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30958 (vectorizable_bswap): Likewise.
30959 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
30960 explicit encoding for the power-of-2 permutes.
30961 (vect_permute_store_chain): Likewise.
30962 (vect_grouped_load_supported): Likewise.
30963 (vect_permute_load_chain): Likewise.
30965 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30967 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
30968 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
30969 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
30970 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
30971 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
30972 (vect_gen_perm_mask_any): Likewise.
30974 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
30976 * int-vector-builder.h: New file.
30977 * vec-perm-indices.h: Include int-vector-builder.h.
30978 (vec_perm_indices): Redefine as an int_vector_builder.
30979 (auto_vec_perm_indices): Delete.
30980 (vec_perm_builder): Redefine as a stand-alone class.
30981 (vec_perm_indices::vec_perm_indices): New function.
30982 (vec_perm_indices::clamp): Likewise.
30983 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
30984 (vec_perm_indices::new_vector): New function.
30985 (vec_perm_indices::new_expanded_vector): Update for new
30986 vec_perm_indices class.
30987 (vec_perm_indices::rotate_inputs): New function.
30988 (vec_perm_indices::all_in_range_p): Operate directly on the
30989 encoded form, without computing elided elements.
30990 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
30991 encoding. Update for new vec_perm_indices class.
30992 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
30993 the given vec_perm_builder.
30994 (expand_vec_perm_var): Update vec_perm_builder constructor.
30995 (expand_mult_highpart): Use vec_perm_builder instead of
30996 auto_vec_perm_indices.
30997 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
30998 vec_perm_indices instead of auto_vec_perm_indices. Use a single
30999 or double series encoding as appropriate.
31000 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
31001 vec_perm_indices instead of auto_vec_perm_indices.
31002 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31003 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31004 (vect_permute_store_chain): Likewise.
31005 (vect_grouped_load_supported): Likewise.
31006 (vect_permute_load_chain): Likewise.
31007 (vect_shift_permute_load_chain): Likewise.
31008 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
31009 (vect_transform_slp_perm_load): Likewise.
31010 (vect_schedule_slp_instance): Likewise.
31011 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31012 (vectorizable_mask_load_store): Likewise.
31013 (vectorizable_bswap): Likewise.
31014 (vectorizable_store): Likewise.
31015 (vectorizable_load): Likewise.
31016 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
31017 vec_perm_indices instead of auto_vec_perm_indices. Use
31018 tree_to_vec_perm_builder to read the vector from a tree.
31019 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
31020 vec_perm_builder instead of a vec_perm_indices.
31021 (have_whole_vector_shift): Use vec_perm_builder and
31022 vec_perm_indices instead of auto_vec_perm_indices. Leave the
31023 truncation to calc_vec_perm_mask_for_shift.
31024 (vect_create_epilog_for_reduction): Likewise.
31025 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
31026 from auto_vec_perm_indices to vec_perm_indices.
31027 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
31028 instead of changing individual elements.
31029 (aarch64_vectorize_vec_perm_const): Use new_vector to install
31030 the vector in d.perm.
31031 * config/arm/arm.c (expand_vec_perm_d::perm): Change
31032 from auto_vec_perm_indices to vec_perm_indices.
31033 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
31034 instead of changing individual elements.
31035 (arm_vectorize_vec_perm_const): Use new_vector to install
31036 the vector in d.perm.
31037 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
31038 Update vec_perm_builder constructor.
31039 (rs6000_expand_interleave): Likewise.
31040 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
31041 (rs6000_expand_interleave): Likewise.
31043 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31045 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
31046 to qimode could truncate the indices.
31047 * optabs.c (expand_vec_perm_var): Likewise.
31049 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31051 * Makefile.in (OBJS): Add vec-perm-indices.o.
31052 * vec-perm-indices.h: New file.
31053 * vec-perm-indices.c: Likewise.
31054 * target.h (vec_perm_indices): Replace with a forward class
31056 (auto_vec_perm_indices): Move to vec-perm-indices.h.
31057 * optabs.h: Include vec-perm-indices.h.
31058 (expand_vec_perm): Delete.
31059 (selector_fits_mode_p, expand_vec_perm_var): Declare.
31060 (expand_vec_perm_const): Declare.
31061 * target.def (vec_perm_const_ok): Replace with...
31062 (vec_perm_const): ...this new hook.
31063 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
31064 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
31065 * doc/tm.texi: Regenerate.
31066 * optabs.def (vec_perm_const): Delete.
31067 * doc/md.texi (vec_perm_const): Likewise.
31068 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
31069 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
31070 expand_vec_perm for constant permutation vectors. Assert that
31071 the mode of variable permutation vectors is the integer equivalent
31072 of the mode that is being permuted.
31073 * optabs-query.h (selector_fits_mode_p): Declare.
31074 * optabs-query.c: Include vec-perm-indices.h.
31075 (selector_fits_mode_p): New function.
31076 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
31077 is defined, instead of checking whether the vec_perm_const_optab
31078 exists. Use targetm.vectorize.vec_perm_const instead of
31079 targetm.vectorize.vec_perm_const_ok. Check whether the indices
31080 fit in the vector mode before using a variable permute.
31081 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
31082 vec_perm_indices instead of an rtx.
31083 (expand_vec_perm): Replace with...
31084 (expand_vec_perm_const): ...this new function. Take the selector
31085 as a vec_perm_indices rather than an rtx. Also take the mode of
31086 the selector. Update call to shift_amt_for_vec_perm_mask.
31087 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
31088 Use vec_perm_indices::new_expanded_vector to expand the original
31089 selector into bytes. Check whether the indices fit in the vector
31090 mode before using a variable permute.
31091 (expand_vec_perm_var): Make global.
31092 (expand_mult_highpart): Use expand_vec_perm_const.
31093 * fold-const.c: Includes vec-perm-indices.h.
31094 * tree-ssa-forwprop.c: Likewise.
31095 * tree-vect-data-refs.c: Likewise.
31096 * tree-vect-generic.c: Likewise.
31097 * tree-vect-loop.c: Likewise.
31098 * tree-vect-slp.c: Likewise.
31099 * tree-vect-stmts.c: Likewise.
31100 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
31102 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
31103 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
31104 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
31105 (aarch64_vectorize_vec_perm_const): ...this new function.
31106 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31107 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31108 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
31109 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
31110 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31111 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31112 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
31114 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
31115 check for NEON modes.
31116 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
31117 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
31118 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
31119 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
31121 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
31122 the old VEC_PERM_CONST conditions.
31123 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
31124 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
31125 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
31126 (ia64_vectorize_vec_perm_const_ok): Merge into...
31127 (ia64_vectorize_vec_perm_const): ...this new function.
31128 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
31129 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
31130 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
31131 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
31132 * config/mips/mips.c (mips_expand_vec_perm_const)
31133 (mips_vectorize_vec_perm_const_ok): Merge into...
31134 (mips_vectorize_vec_perm_const): ...this new function.
31135 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
31136 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
31137 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
31138 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
31139 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
31140 (rs6000_expand_vec_perm_const): Delete.
31141 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
31143 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31144 (altivec_expand_vec_perm_const_le): Take each operand individually.
31145 Operate on constant selectors rather than rtxes.
31146 (altivec_expand_vec_perm_const): Likewise. Update call to
31147 altivec_expand_vec_perm_const_le.
31148 (rs6000_expand_vec_perm_const): Delete.
31149 (rs6000_vectorize_vec_perm_const_ok): Delete.
31150 (rs6000_vectorize_vec_perm_const): New function.
31151 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
31152 an element count and rtx array.
31153 (rs6000_expand_extract_even): Update call accordingly.
31154 (rs6000_expand_interleave): Likewise.
31155 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
31156 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
31157 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
31158 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
31159 (rs6000_expand_vec_perm_const): Delete.
31160 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31161 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31162 (altivec_expand_vec_perm_const_le): Take each operand individually.
31163 Operate on constant selectors rather than rtxes.
31164 (altivec_expand_vec_perm_const): Likewise. Update call to
31165 altivec_expand_vec_perm_const_le.
31166 (rs6000_expand_vec_perm_const): Delete.
31167 (rs6000_vectorize_vec_perm_const_ok): Delete.
31168 (rs6000_vectorize_vec_perm_const): New function. Remove stray
31169 reference to the SPE evmerge intructions.
31170 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
31171 an element count and rtx array.
31172 (rs6000_expand_extract_even): Update call accordingly.
31173 (rs6000_expand_interleave): Likewise.
31174 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
31175 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
31177 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31179 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31181 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
31182 vector mode and that that mode matches the mode of the data
31184 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
31185 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
31186 directly using expand_vec_perm_1 when forcing selectors into
31188 (expand_vec_perm_var): New function, split out from expand_vec_perm.
31190 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31192 * optabs-query.h (can_vec_perm_p): Delete.
31193 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
31194 * optabs-query.c (can_vec_perm_p): Split into...
31195 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
31196 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
31197 particular selector is valid.
31198 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31199 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31200 (vect_grouped_load_supported): Likewise.
31201 (vect_shift_permute_load_chain): Likewise.
31202 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
31203 (vect_transform_slp_perm_load): Likewise.
31204 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31205 (vectorizable_bswap): Likewise.
31206 (vect_gen_perm_mask_checked): Likewise.
31207 * fold-const.c (fold_ternary_loc): Likewise. Don't take
31208 implementations of variable permutation vectors into account
31209 when deciding which selector to use.
31210 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
31211 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
31212 with a false third argument.
31213 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
31214 to test whether the constant selector is valid and can_vec_perm_var_p
31215 to test whether a variable selector is valid.
31217 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31219 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
31220 * optabs-query.c (can_vec_perm_p): Likewise.
31221 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
31222 instead of vec_perm_indices.
31223 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
31224 (vect_gen_perm_mask_checked): Likewise,
31225 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
31226 (vect_gen_perm_mask_checked): Likewise,
31228 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31230 * optabs-query.h (qimode_for_vec_perm): Declare.
31231 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
31232 (qimode_for_vec_perm): ...this new function.
31233 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
31235 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
31237 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
31238 does not have a conditional at the top.
31240 2018-01-02 Richard Biener <rguenther@suse.de>
31242 * ipa-inline.c (big_speedup_p): Fix expression.
31244 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
31247 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
31250 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
31254 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
31255 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
31256 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
31257 cond_taken_branch_cost 3->4.
31259 2018-01-01 Jakub Jelinek <jakub@redhat.com>
31261 PR tree-optimization/83581
31262 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
31263 TODO_cleanup_cfg if any changes have been made.
31265 PR middle-end/83608
31266 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
31267 convert_modes if target mode has the right side, but different mode
31270 PR middle-end/83609
31271 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
31272 last argument when extracting from CONCAT. If either from_real or
31273 from_imag is NULL, use expansion through memory. If result is not
31274 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
31275 the parts directly to inner mode, if even that fails, use expansion
31278 PR middle-end/83623
31279 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
31280 check for bswap in mode rather than HImode and use that in expand_unop
31283 Copyright (C) 2018 Free Software Foundation, Inc.
31285 Copying and distribution of this file, with or without modification,
31286 are permitted in any medium without royalty provided the copyright
31287 notice and this notice are preserved.