2018-11-09 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blob693ac972e5ffe3e1a81c5af18fa7dec4525b51c6
1 2018-11-09  Richard Biener  <rguenther@suse.de>
3         PR tree-optimization/87953
4         * tree-vect-loop.c (vectorizable_reduction): For analysis
5         always pass ops[0] to vectorizable_condition.
7 2018-11-09  Stafford Horne  <shorne@gmail.com>
8             Richard Henderson  <rth@twiddle.net>
9             Joel Sherrill  <joel@rtems.org>
11         * common/config/or1k/or1k-common.c: New file.
12         * config/or1k/*: New.
13         * config.gcc (or1k*-*-*): New.
14         * configure.ac (or1k*-*-*): New test for openrisc tls.
15         * configure: Regenerated.
16         * doc/install.texi: Document OpenRISC triplets.
17         * doc/invoke.texi: Document OpenRISC arguments.
18         * doc/md.texi: Document OpenRISC.
20 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
22         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
23         (arm7tdmi-s): Delete CPU.
24         (arm710t): Add aliases for arm720t and arm740t.
25         (arm720t, arm740t): Delete CPUs.
26         (arm920t): Add aliases for arm920, arm922t and arm940t.
27         (arm920, arm922t, arm940t): Delete CPUs.
28         (arm10tdmi): Add alias for arm1020t.
29         (arm1020t): Delete CPU.
30         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
31         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
32         (arm10e): Add aliases for arm1020e and arm1022e.
33         (arm1020e, arm1022e): Delete CPU.
34         * config/arm/arm.md (generic_sched): Remove entries that are now
35         handled by aliases.
36         (generic_vfp): Likewise.
37         * config/arm/arm1020e.md: Simplify tuning selection based on alias
38         changes.
39         * config/arm/arm-tune.md: Regenerated.
40         * config/arm/arm-tables.opt: Regenerated.
42 2018-11-09  Richard Biener  <rguenther@suse.de>
44         PR tree-optimization/87621
45         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
46         op with only phi inputs.
47         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
48         (ch_base::copy_headers): Run CSE on copied loop headers.
49         (pass_ch_vect::process_loop_p): Simplify.
51 2018-11-09  Alexandre Oliva <oliva@adacore.com>
53         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
54         for non-w64 x86_64 biarch.
56 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
58         PR rtl-optimization/86438
59         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
60         of in_b for the compare if in_b is SET_DEST.
62         PR target/87793
63         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
64         non-toplevel UNSPEC.
66 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
68         * tree-vrp.c (value_range::check): Do not access internals
69         directly.
70         (value_range::singleton_p): Same.
71         (value_range::type): Same.
72         (vrp_finalize): Use value_range API.
74 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
76         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
78 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
80         * vr-values.c (vr_values::get_value_range): Use value_range API
81         instead of piecing together ranges.
82         (vr_values::update_value_range): Same.
84 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
86         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
87         of performing ad-hoc calculations.
88         * tree-ssanames.c (set_range_info): New overloaded function
89         accepting value_range &.
90         (get_range_info): Same.
91         * tree-ssanames.h (set_range_info_raw): Remove.
92         (set_range_info): New prototype.
93         (get_range_info): Same.
94         * tree-vrp.h (value_range::null_p): Rename to zero_p.
95         * tree-vrp.c (value_range::null_p): Same.
97 2018-11-09  Jan Hubicka  <jh@suse.cz>
99         * tree.c (fld_type_variant_equal_p): Test user align flag.
100         (flt_type_variant): Copy user align flag.
101         (fld_incomplete_type_of): Clear it.
103 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
105         * config/arm/neon.md (div<mode>3): New pattern.
107 2018-11-08  Andi Kleen  <ak@linux.intel.com>
109         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
110         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
111         (ix86_handle_option): Handle OPT_mptwrite.
112         * config/i386/cpuid.h (bit_PTWRITE): Add.
113         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
114         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
115         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
116         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
117         (ix86_option_override_internal): Handle PTA_PTWRITE.
118         (ix86_valid_target_attribute_inner_p): Define ptwrite.
119         (def_builtin2): Force UINT64 to be 64bit only.
120         * config/i386/i386.h (TARGET_PTWRITE): Add.
121         (TARGET_PTWRITE_P): Add.
122         (PTA_PTWRITE): Add.
123         * config/i386/i386.md: Define ptwrite.
124         * config/i386/i386.opt: Add -mptwrite.
125         * config/i386/immintrin.h (_ptwrite64): Add.
126         (_ptwrite32): Add
127         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
128         * doc/invoke.texi: Document -mptwrite.
130 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
132         PR rtl-optimization/87600
133         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
134         * lra-constraints.c (process_alt_operands): Skip illegal hard
135         register usage.  Prefer reloading non hard register operands.
137 gcc/testsuite/
138         PR rtl-optimization/87600
139         * gcc.dg/pr87600.h: New file.
140         * gcc.dg/pr87600-1.c: New test.
141         * gcc.dg/pr87600-2.c: Likewise.
143 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
145         PR other/36572
146         * doc/invoke.texi (Optimize Options): Clarify default behavior
147         for -fno-sched-interblock and -fno-sched-spec.
149 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
151         * collect2.c (linker_select):  Add USE_LLD_LD.
152         (ld_suffixes): Add ld.lld.
153         (main): Handle -fuse-ld=lld.
154         * common.opt (-fuse-ld=lld): New option.
155         * doc/invoke.texi (-fuse-ld=lld): Document.
156         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
158 2018-11-08  Paul Koning  <ni1d@arrl.net>
160         * config/pdp11/constraints.md: Add "Z" series constraints for use
161         with pre-dec and post-inc addressing.
162         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
163         (pdp11_expand_operands): Add int argument (word count).
164         (pdp11_sp_frame_offset): Delete.
165         (pdp11_cmp_length): New function.
166         (pushpop_regeq): New function.
167         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
168         Add hook.
169         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
170         frame layout.
171         (pdp11_initial_elimination_offset): Ditto.
172         (pdp11_expand_operands): Add word count argument.  Bugfixes.
173         (output_move_multiple): Change how pointer adjustment is done.
174         (pdp11_gen_int_label): Correct format.
175         (output_ascii): Ditto.
176         (pdp11_asm_output_var): Add code for DEC assembler case.
177         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
178         value.
179         (legitimate_const_double_p): Ditto.
180         (pdp11_register_move_cost): Adjust for new register classes.
181         (pdp11_regno_reg_class): Ditto.
182         (expand_block_move): Delete.
183         (pushpop_regeq): New function.
184         (pdp11_legitimate_address_p): Bugfix in check for constant
185         offset.
186         (pdp11_sp_frame_offset): Delete.
187         (pdp11_reg_save_size): New helper function for new frame layout.
188         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
189         (pdp11_expand_shift): Bugfix in check for constant shift count.
190         (pdp11_shift_length): Ditto.
191         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
192         (pdp11_cmp_length): New function.
193         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
194         some compile options.
195         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
196         (CALL_USED_REGISTERS): Ditto.
197         (ELIMINABLE_REGS): Ditto.
198         (REGISTER_NAMES): Ditto.
199         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
200         constraints.
201         (REG_CLASS_NAMES): Ditto.
202         (REG_CLASS_CONTENTS): Ditto.  Also remove
203         HARD_FRAME_POINTER_REGNUM.
204         (CPU_REG_CLASS): New macro.
205         (CLASS_MAX_NREGS): Adjust for new register classes.
206         (FUNCTION_PROFILER): Make no-op.
207         (may_call_alloca): Remove unused declaration.
208         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
209         (ASM_OUTPUT_SKIP): Fix format.
210         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
211         (HARD_FRAME_POINTER_REGNUM): Remove.
212         (return): Delete.
213         (*rts): Rename.  Remove epilogue related checks.
214         (cmpsi, cmpdi): New insn.
215         (cbranch<mode>4): Change to apply to SI and DI modes as well.
216         (mov<mode>): Change constraints to enforce that push/pop
217         destination cannot use the same register as source.
218         (*mov<mode><cc_cc>): Ditto.
219         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
220         at assembly output rather than as RTL expander.
221         (zero_extendqihi2): Bugfix in check for same registers.
222         (adddi3_nocc): Bugfix in check for constant operand.
223         (addsi3_nocc): Ditto.
224         (subdi3_nocc): Ditto.
225         (subsi3_nocc): Ditto.
226         (negdi2_nocc): Copy input to pdp11_expand_operands.
227         (negsi2_nocc): Ditto.
228         (bswap2_nocc): Ditto.
229         * config/pdp11/pdp11.opt (mlra): Fix documentation.
230         * config/pdp11/t-pdp11: Use -Os.
232 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
234         * config/arm/parsecpu.awk (/alias/): New parsing rule.
235         (/begin cpu/): Check that the cpu name hasn't been previously defined.
236         (gen_comm_data): Print out CPU alias tables.
237         (check_cpu): Match aliases when checking the CPU name.
238         * config/arm/arm-protos.h (cpu_alias): New structure.
239         (cpu_option): Add entry for aliases.
240         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
241         strongarm1100 and strongarm1110.
242         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
243         (config/arm/arm-generic.md): Remove redundant references to
244         strongarm110, strongarm1100 and strongarm1110.
245         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
246         Scan aliases for additional hints.
247         (arm_parse_cpu_option_name): Also match a cpu name against the list
248         of aliases.
249         * config/arm/arm-tables.opt: Regenerated.
250         * config/arm/arm-tune.md: Regenerated.
252 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
254         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
255         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
256         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
257         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
258         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
259         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
260         * gengtype.c (open_base_files): Add omp-general.h.
261         * gimple.c (gimple_build_omp_critical):
262         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
263         gimple_omp_taskgroup_set_clauses.
264         (gimple_build_omp_atomic_load): Add mo argument, call
265         gimple_omp_atomic_set_memory_order.
266         (gimple_build_omp_atomic_store): Likewise.
267         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
268         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
269         instead of GSS_OMP.
270         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
271         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
272         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
273         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
274         different value for GF_OMP_ATOMIC_NEED_VALUE.
275         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
276         comments.
277         (struct gimple_statement_omp_single_layout): And remove here.
278         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
279         than gimple_statement_omp_single_layout.
280         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
281         GIMPLE_OMP_TEAMS.
282         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
283         (gimple_omp_subcode): Formatting fix.
284         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
285         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
286         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
287         gimple_omp_teams_host, gimple_omp_teams_set_host,
288         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
289         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
290         gimple_omp_taskgroup_set_clauses): New inline functions.
291         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
292         (gimple_build_omp_atomic_store): Likewise.
293         (gimple_omp_atomic_seq_cst_p): Remove.
294         (gimple_omp_atomic_memory_order): New function.
295         (gimple_omp_atomic_set_seq_cst): Remove.
296         (gimple_omp_atomic_set_memory_order): New function.
297         (gimple_build_omp_taskgroup): Add clauses argument.
298         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
299         (dump_gimple_omp_task): Print taskwait with depend clauses.
300         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
301         dump_omp_atomic_memory_order.
302         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
303         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
304         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
305         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
306         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
307         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
308         ORT_UNTIED_TASKLOOP enumerators.
309         (enum gimplify_defaultmap_kind): New.
310         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
311         target_map_pointers_as_0len_arrays members, add defaultmap.
312         (new_omp_context): Initialize defaultmap member.
313         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
314         (maybe_fold_stmt): Don't fold even in host teams regions.
315         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
316         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
317         ctx->omp_firstprivatize_variable.
318         (omp_add_variable): Don't add private/firstprivate for VLAs in
319         ORT_TASKGROUP.
320         (omp_default_clause): Print "taskloop" rather than "task" if
321         ORT_*TASKLOOP.
322         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
323         Handle new defaultmap clause kinds.
324         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
325         iterator to be lastprivate or private.  Fix up diagnostics if linear
326         is used on collapse>1 simd iterator.
327         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
328         (gimplify_omp_depend): New function.
329         (gimplify_scan_omp_clauses): Add shared clause on parallel for
330         combined parallel master taskloop{, simd} if taskloop has
331         firstprivate, lastprivate or reduction clause.  Handle
332         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
333         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
334         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
335         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
336         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
337         lastprivate.
338         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
339         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.  
340         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
341         OMP_CLAUSE_{TASK,IN}_REDUCTION.
342         (gimplify_omp_task): Handle taskwait with depend clauses.
343         (gimplify_omp_for): Add shared clause on parallel for combined
344         parallel master taskloop{, simd} if taskloop has firstprivate,
345         lastprivate or reduction clause.  Use ORT_TASKLOOP or
346         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
347         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
348         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
349         __for_end and __for_range temporaries on OMP_PARALLEL for
350         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
351         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
352         sandwiched in between two taskloops.
353         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
354         instead of ctx->omp_firstprivatize_variable.
355         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
356         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
357         host teams, use gimplify_and_return_first etc. for body like
358         for target or target data constructs, and at the end call
359         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
360         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
361         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
362         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
363         gimple_omp_atomic_set_seq_cst calls.
364         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
365         case, handle taskgroup clauses.
366         * lto-streamer-out.c (hash_tree): Handle
367         OMP_CLAUSE_{TASK,IN}_REDUCTION.
368         * Makefile.in (GTFILES): Add omp-general.h.
369         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
370         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
371         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
372         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
373         BUILT_IN_GOMP_LOOP_DOACROSS_START,
374         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
375         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
376         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
377         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
378         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
379         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
380         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
381         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
382         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
383         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
384         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
385         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
386         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
387         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
388         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
389         * omp-expand.c (workshare_safe_to_combine_p): Return false for
390         non-worksharing loops.
391         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
392         (determine_parallel_type): Don't combine parallel with worksharing
393         which has _reductemp_ clause.
394         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
395         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
396         if there is nonmonotonic modifier or if there is no modifier and no
397         ordered clause.  For dynamic and guided schedule without monotonic
398         and nonmonotonic modifier, default to nonmonotonic.
399         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
400         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
401         task reductions.
402         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
403         there are any reduction clauses.
404         (expand_taskwait_call): New function.
405         (expand_teams_call): New function.
406         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
407         expand_teams_call for it.  Formatting fix.  Handle taskwait with
408         depend clauses.
409         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
410         of worksharing loops with task reductions.
411         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
412         expansion of worksharing loops with task reductions.
413         (expand_omp_sections): Handle expansion of sections with task
414         reductions.
415         (expand_omp_synch): For host teams call expand_omp_taskreg.
416         (omp_memory_order_to_memmodel): New function.
417         (expand_omp_atomic_load, expand_omp_atomic_store,
418         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
419         instead of gimple_omp_atomic_seq_cst_p.
420         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
421         depend clauses as a standalone directive.
422         * omp-general.c (enum omp_requires): New variable.
423         (omp_extract_for_data): Initialize have_reductemp member.  Allow
424         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
425         GT_EXPR loops depending on incr sign.  Formatting fixes.
426         * omp-general.h (struct omp_for_data): Add have_reductemp member.
427         (enum omp_requires): New enum.
428         (omp_requires_mask): Declare.
429         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
430         Fix comment typos.
431         * omp-low.c (struct omp_context): Add task_reductions and
432         task_reduction_map fields.
433         (is_host_teams_ctx): New function.
434         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
435         (use_pointer_for_field): Use is_global_var instead of
436         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
437         in outer contexts.
438         (build_outer_var_ref): Ignore taskgroup outer contexts.
439         (delete_omp_context): Release task_reductions and task_reduction_map.
440         (scan_sharing_clauses): Don't add any fields for reduction clause on
441         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
442         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
443         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
444         Handle OMP_CLAUSE_NONTEMPORAL.
445         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
446         needed.
447         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
448         clauses with task modifier.
449         (scan_omp_task): Handle taskwait with depend clauses.
450         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
451         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
452         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
453         (scan_omp_for): Fix comment formatting.
454         (scan_omp_teams): Handle host teams constructs.
455         (check_omp_nesting_restrictions): Allow teams with no outer
456         OpenMP context.  Adjust diagnostics for teams strictly nested into
457         some explicit OpenMP construct other than target.  Allow OpenMP atomics
458         inside of simd regions.
459         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
460         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
461         taskreg_nesting_level while scanning host teams construct.
462         (task_reduction_read): New function.
463         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
464         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
465         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
466         second argument create_tmp_var if it is NULL.  Don't ignore shared
467         clauses in is_host_teams_ctx contexts.  Handle
468         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
469         clauses.
470         (lower_reduction_clauses): Ignore reduction clauses with task
471         modifier.  Remove second argument create_tmp_var if it is NULL.
472         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
473         (lower_send_clauses): Ignore reduction clauses with task modifier.
474         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
475         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
476         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
477         rely that it is the last stmt in body so far.  Ignore outer taskgroup
478         contexts.
479         (omp_task_reductions_find_first, omp_task_reduction_iterate,
480         lower_omp_task_reductions): New functions.
481         (lower_omp_sections): Handle reduction clauses with taskgroup
482         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
483         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
484         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
485         modifiers.
486         (lower_omp_taskgroup): Handle taskgroup reductions.
487         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
488         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
489         (lower_depend_clauses): If there are any
490         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
491         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
492         seen, assume lowering is done already and return early.  Set kind
493         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
494         (lower_omp_taskreg): Handle reduction clauses with task modifier on
495         parallel construct.  Handle reduction clause on taskloop construct.
496         Handle taskwait with depend clauses.
497         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
498         for host teams constructs.
499         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
500         nontemporal and _reductemp_ clause entries.
501         (omp_clause_code_name): Likewise.
502         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
503         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
504         * tree-core.h (enum omp_clause_code): Add
505         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
506         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
507         (struct tree_base): Add omp_atomic_memory_order field into union.
508         Remove OMP_ATOMIC_SEQ_CST comment.
509         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
510         and OMP_CLAUSE_DEPEND_DEPOBJ.
511         (struct tree_omp_clause): Add subcode.defaultmap_kind.
512         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
513         OpenMP constructs with body and clauses operands.
514         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
515         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
516         (OMP_TASKGROUP_CLAUSES): Define.
517         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
518         OMP_CLAUSE__LOOPTEMP_.
519         (OMP_ATOMIC_SEQ_CST): Remove.
520         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
521         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
522         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
523         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
524         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
525         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
526         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
527         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
528         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
529         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
530         Define.
531         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
532         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
533         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
534         (convert_local_omp_clauses): Likewise.  Remove useless test.
535         * tree-parloops.c (create_call_for_reduction_1): Pass
536         OMP_MEMORY_ORDER_RELAXED as new argument to
537         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
538         * tree-pretty-print.c (dump_omp_iterators): New function.
539         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
540         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
541         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
542         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
543         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
544         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
545         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
546         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
547         (dump_omp_atomic_memory_order): New function.
548         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
549         taskwait with depend clauses.
550         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
551         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
552         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
553         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
554         write_ts_omp_clause_tree_pointers): Likewise.
556 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
558         PR ipa/86395
559         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
560         "-missed", "-note", and "-all" sub-options.
561         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
562         API.
563         (can_early_inline_edge_p): Likewise.
564         (want_early_inline_function_p): Likewise.
565         (want_inline_self_recursive_call_p): Likewise.
566         (recursive_inlining): Likewise.
567         (inline_small_functions): Likewise.
568         (flatten_function): Likewise.
569         (ipa_inline): Likewise.
570         (inline_always_inline_functions): Likewise.
571         (early_inline_small_functions): Likewise.
572         (early_inliner): Likewise.
573         * tree-inline.c (expand_call_inline): Likewise.
575 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
577         * pretty-print.c (pp_format): Handle %f.
578         (selftest::test_pp_format): Add test of %f.
579         * pretty-print.h (pp_double): New macro.
581 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
583         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
584         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
585         (dump_pretty_printer::decode_format): Implement "%C" for
586         cgraph_node *.
587         (selftest::test_capture_of_dump_calls): Rename "where" to
588         "stmt_loc".  Convert test_decl to a function decl and set its
589         location.  Add a symbol_table_test RAII instance and a
590         cgraph_node, using it to test "%C" and dump_symtab_node.
592 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
594         PR middle-end/87916
595         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
597 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
599         * cgraph.c: Include "selftest.h".
600         (saved_symtab): New variable.
601         (selftest::symbol_table_test::symbol_table_test): New ctor.
602         (selftest::symbol_table_test::~symbol_table_test): New dtor.
603         (selftest::test_symbol_table_test): New test.
604         (selftest::cgraph_c_tests): New.
605         * cgraph.h (saved_symtab): New decl.
606         (selftest::symbol_table_test): New class.
607         * selftest-run-tests.c (selftest::run_tests): Call
608         selftest::cgraph_c_tests.
609         * selftest.h (selftest::cgraph_c_tests): New decl.
611 2018-11-08  Richard Biener  <rguenther@suse.de>
613         * tree-data-ref.h (lambda_int): New typedef.
614         (lambda_vector_gcd): Adjust.
615         (lambda_vector_new): Likewise.
616         (lambda_matrix_new): Likewise.
617         * tree-data-ref.c  (print_lambda_vector): Adjust.
619 2018-11-08  Richard Biener  <rguenther@suse.de>
621         PR tree-optimization/87929
622         * tree-complex.c (expand_complex_comparison): Clean EH.
624 2018-11-08  Martin Liska  <mliska@suse.cz>
626         * doc/extend.texi: Reword.
627         * predict.c (expr_expected_value_1): Likewise.
629 2018-11-08  Richard Biener  <rguenther@suse.de>
631         PR tree-optimization/87913
632         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
633         of extreme values to ordered comparisons.
635 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
637         PR middle-end/42726
638         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
640 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
642         * doc/invoke.texi: Remove leading dash from @opindex entries
643         throughout the file.
645 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
647         PR driver/80828
648         * doc/invoke.texi (Option Summary): Add -e and --entry.
649         (Link Options): Likewise.
651 2018-11-07  Nathan Sidwell  <nathan@acm.org>
653         PR 87926
654         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
655         --disable-checking bootstrap.
657 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
659         * configure: Regenerated.
661 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
663         PR c/87691
664         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
665         to the mode of the widest field iff the widest field has mode class
666         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
667         reference.
669 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
671         * common.opt: Add -Wattribute-warning.
672         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
673         * expr.c (expand_expr_real_1): Add new attribute to warning_at
674         call to allow user configure behavior of "warning" attribute.
676 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
678         * target.def: Put @: after every vs., e.g., and i.e. where it is
679         followed by whitespace.
680         * doc/extend.texi: Ditto.
681         * doc/fragments.texi: Ditto.
682         * doc/gimple.texi: Ditto.
683         * doc/implement-c.texi: Ditto.
684         * doc/install.texi: Ditto.
685         * doc/invoke.texi: Ditto.
686         * doc/md.texi: Ditto.
687         * doc/plugins.texi: Ditto.
688         * doc/rtl.texi: Ditto.
689         * doc/sourcebuild.texi: Ditto.
690         * doc/tm.texi.in: Ditto.
691         * doc/ux.texi: Ditto.
692         * doc/tm.texi: Regenerate.
694 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
696         * config/arm/arm-cpus.in (ares): New entry.
697         * config/arm/arm-tables.opt: Regenerate.
698         * config/arm/arm-tune.md: Likewise.
699         * doc/invoke.texi (ARM Options): Document ares.
701 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
703         * config/aarch64/aarch64-cores.def (ares): Define.
704         * config/aarch64/aarch64-tune.md: Regenerate.
705         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
707 2018-11-07  Jan Hubicka  <jh@suse.cz>
709         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
710         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
711         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
712         TYPE_DECL.
714 2018-11-07  Richard Biener  <rguenther@suse.de>
716         PR tree-optimization/87914
717         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
718         of nested cycles.
719         (vectorizable_reduction): Handle shifts and rotates by dispatching
720         to vectorizable_shift.
721         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
722         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
723         def cases.
724         (vectorizable_shift): Export and handle being called as
725         vect_nested_cycle.
726         (vect_analyze_stmt): Call vectorizable_shift after
727         vectorizable_reduction.
728         * tree-vectorizer.h (vectorizable_shift): Declare.
730 2018-11-07  Jan Hubicka  <jh@suse.cz>
732         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
733         than const decls in TREE_VALUE of enum.
734         (dump_type_inheritance_graph): Improve duplicate dumping.
735         (free_enum_values): New.
736         (build_type_inheritance_graph): Use it.
737         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
738         which are not main variants or not ODR types.
739         (verify_type_variant): Expect variants to have no TYPE_VALUES.
741 2018-11-07  Richard Biener  <rguenther@suse.de>
743         * ipa-inline.c (want_inline_small_function_p): Compute
744         big_speedup_p lazily and last.
746 2018-11-07  Jan Hubicka  <jh@suse.cz>
748         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
749         building incomplete variant of complete type.
750         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
751         variant of complete type.
753 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
755         * config/mips/mips.c: Fix typo in documentation of
756         mips_loongson_ext2_prefetch_cookie.
757         (mips_option_override): fix brain twister logical.
758         * config/mips/mips.h: Fix typo in documentation of
759         ISA_HAS_CTZ_CTO and define pattern.
760         * config/mips/mips.md (prefetch): Hoist EXT2 above
761         the 2EF/EXT block.
762         (prefetch_indexed): Hoist EXT2 above the EXT block.
764 2018-11-07  Jan Hubicka  <jh@suse.cz>
766         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
767         return and parameter types of function and method types.
768         (free_lang_data_in_cgraph): Update.
770 2018-11-07  Martin Liska  <mliska@suse.cz>
772         PR rtl-optimization/87868
773         * postreload-gcse.c (eliminate_partially_redundant_load): Set
774         threshold to max_count if we would overflow.
775         * profile-count.h: Make max_count a public constant.
777 2018-11-07  Martin Liska  <mliska@suse.cz>
779         * mem-stats.h: Fix GNU coding style.
781 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
783         * config/mips/gs264e.md: New.
784         * config/mips/mips-cpus.def: Define gs264e.
785         * config/mips/mips-tables.opt: Regenerate.
786         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
787         gs264e.
788         (mips_issue_rate): Add support for gs264e.
789         (mips_multipass_dfa_lookahead): Likewise.
790         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
791         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
792         (MIPS_ASE_MSA_SPEC): New.
793         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
794         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
795         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
796         * config/mips/mips.md: Include gs264e.md.
797         (processor): Add gs264e.
798         * config/mips/mips.opt (MSA): Use Mask instead of Var.
799         * doc/invoke.texi: Add gs264e to supported architectures.
801 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
803         * config/mips/gs464e.md: New.
804         * config/mips/mips-cpus.def: Define gs464e.
805         * config/mips/mips-tables.opt: Regenerate.
806         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
807         gs464e.
808         (mips_issue_rate): Add support for gs464e.
809         (mips_multipass_dfa_lookahead): Likewise.
810         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
811         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
812         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
813         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
814         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
815         * config/mips/mips.md: Include gs464e.md.
816         (processor): Add gs464e.
817         * doc/invoke.texi: Add gs464e to supported architectures.
819 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
821         * config/mips/loongson3a.md: Rename to ...
822         * config/mips/gs464.md: ... here.
823         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
824         as an alias of gs464 processor.
825         * config/mips/mips-tables.opt: Regenerate.
826         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
827         instead of PROCESSOR_LOONGSON_3A.
828         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
829         TUNE_LOONGSON_3A.
830         (mips_option_override): Enable MMI and EXT for gs464.
831         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
832         Rename TUNE_LOONGSON_3A to TUNE_GS464.
833         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
834         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
835         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
836         TARGET_LOONGSON_3A.
837         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
838         (processor): Add gs464;
839         * doc/invoke.texi: Add gs464 to supported architectures.
841 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
843         * config/mips/mips-protos.h
844         (mips_loongson_ext2_prefetch_cookie): New prototype.
845         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
846         (mips_option_override): Enable TARGET_LOONGSON_EXT when
847         TARGET_LOONGSON_EXT2 is true.
848         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
849         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
850         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
851         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
852         TARGET_LOONGSON_EXT2.
853         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
854         (define_insn "ctz<mode>2"): New insn pattern.
855         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
856         (define_insn "prefetch_indexed_<mode>"): Include
857         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
858         * config/mips/mips.opt (-mloongson-ext2): Add option.
859         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
861 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
863         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
864         __mips_loongson_ext.
865         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
866         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
867         -mloongson-ext.
868         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
869         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
870         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
871         instead of TARGET_LOONGSON_3A.
872         * config/mips/mips.opt (-mloongson-ext): Add option.
873         * gcc/doc/invoke.texi (-mloongson-ext): Document.
875 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
877         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
878         * config/mips/loongson.md: Move to ...
879         * config/mips/loongson-mmi.md: here; Adjustment.
880         * config/mips/loongson.h: Move to ...
881         State as deprecated. Include loongson-mmiintrin.h for back
882         compatibility and warning.
883         * config/mips/loongson-mmiintrin.h: ... here.
884         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
885         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
886         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
887         (mips_option_override): Make sure MMI use hard float;
888         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
889         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
890         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
891         TARGET_LOONGSON_VECTORS.
892         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
893         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
894         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
895         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
896         -mloongson-mmi.
897         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
898         TARGET_LOONGSON_VECTORS.
899         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
900         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
901         (Loongson MMI patterns): Include loongson-mmi.md instead of
902         loongson.md.
903         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
904         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
906 2018-11-07  Richard Biener  <rguenther@suse.de>
908         PR lto/87906
909         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
910         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
912 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
914         PR rtl-optimization/87874
915         * lra.c (lra_substitute_pseudo): Do not create a subreg for
916         const wide ints.
918 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
920         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
921         if not in indexed or indirect form.
922         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
923         (bswap<mode>2_store): Ditto.
925 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
927         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
928         the UNSPEC.
930 2018-11-06  Richard Biener  <rguenther@suse.de>
932         PR tree-optimization/86850
933         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
934         instead of src.m_vec.
936 2018-11-06  Jan Hubicka  <jh@suse.cz>
938         * tree.c (fld_simplified_type_name): Break out form ...
939         (free_lang_data_in_type): ... here.
940         (fld_type_variant_equal_p): Use it.
942 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
944         * config/default-d.c: Include memmodel.h.
946         * config/sol2-d.c: New file.
947         * config/t-sol2 (sol2-d.o): New rule.
948         * config.gcc <*-*-solaris2*>: Set d_target_objs,
949         target_has_targetdm.
951 2018-11-06  Jan Hubicka  <jh@suse.cz>
953         * tree.c (fld_type_variant): Also copy alignment; be sure that
954         new variant is equal.
956 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
958         PR target/87762
959         * config/s390/s390.md: Add relative_long attribute.
961 2018-11-06  Jan Hubicka  <jh@suse.cz>
963         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
964         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
965         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
966         stream TYPE_NEEDS_CONSTRUCTING.
967         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
968         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
970 2018-11-06  Richard Biener  <rguenther@suse.de>
972         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
973         dump-scope ...
974         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
976 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
978         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
979         REG_CLASS_CONTENTS[GEN_REGS].
980         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
982 2018-11-06  Jan Hubicka  <jh@suse.cz>
984          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
986 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
988         PR sanitizer/80953
989         * config/sol2.h (ASAN_CC1_SPEC): Define.
990         (LD_WHOLE_ARCHIVE_OPTION): Define.
991         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
992         (ASAN_REJECT_SPEC): Provide default.
993         (LIBASAN_EARLY_SPEC): Define.
994         (LIBTSAN_EARLY_SPEC): Define.
995         (LIBLSAN_EARLY_SPEC): Define.
996         * config/i386/sol2.h (CC1_SPEC): Redefine.
997         (ASAN_REJECT_SPEC): Define.
999         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
1000         (TARGET_ASAN_SHADOW_OFFSET): Define.
1001         (sparc_asan_shadow_offset): New function.
1002         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
1003         (ASAN_REJECT_SPEC): Define.
1005 2018-11-06  Jan Hubicka  <jh@suse.cz>
1007         * tree.c (fld_type_variant): Copy canonical type.
1008         (fld_incomplete_type_of): Check that canonical types looks sane;
1009         copy canonical type.
1010         (verify_type): Accept when incomplete type has complete canonical type.
1012 2018-11-06  Jan Hubicka  <jh@suse.cz>
1014         * tree.c (free_lang_data): Reset overwite_assembler_name,
1015         print_xnode, print_decl, print_type and print_identifier of
1016         langhooks.
1018 2018-11-06  Richard Biener  <rguenther@suse.de>
1020         PR tree-optimization/87889
1021         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
1022         Do nothing if old and new arg are the same
1024 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
1026         PR target/87723
1027         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
1028         attributes for operands 3 and 4.
1030 2018-11-06  Richard Biener  <rguenther@suse.de>
1032         PR middle-end/18041
1033         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
1034         matching bitfield insertion.
1036 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
1038         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
1039         (reg_next_debug_use): New.
1040         (attempt_change): Propagate adjusted expression into affected
1041         debug insns.
1042         (merge_in_block): Track uses in debug insns.
1043         (pass_inc_dec::execute): Allocate and release
1044         reg_next_debug_use.
1046 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
1048         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
1049         (_mm512_fixupimm_round_pd): Update parameters and builtin.
1050         (_mm512_maskz_fixupimm_round_pd): Ditto.
1051         (_mm512_fixupimm_round_ps): Ditto.
1052         (_mm512_maskz_fixupimm_round_ps): Ditto.
1053         (_mm_fixupimm_round_sd): Ditto.
1054         (_mm_maskz_fixupimm_round_sd): Ditto.
1055         (_mm_fixupimm_round_ss): Ditto.
1056         (_mm_maskz_fixupimm_round_ss): Ditto.
1057         (_mm512_fixupimm_pd): Ditto.
1058         (_mm512_maskz_fixupimm_pd): Ditto.
1059         (_mm512_fixupimm_ps): Ditto.
1060         (_mm512_maskz_fixupimm_ps): Ditto.
1061         (_mm_fixupimm_sd): Ditto.
1062         (_mm_maskz_fixupimm_sd): Ditto.
1063         (_mm_fixupimm_ss): Ditto.
1064         (_mm_maskz_fixupimm_ss): Ditto.
1065         (_mm512_mask_fixupimm_round_pd): Update builtin.
1066         (_mm512_mask_fixupimm_round_ps): Ditto.
1067         (_mm_mask_fixupimm_round_sd): Ditto.
1068         (_mm_mask_fixupimm_round_ss): Ditto.
1069         (_mm512_mask_fixupimm_pd): Ditto.
1070         (_mm512_mask_fixupimm_ps): Ditto.
1071         (_mm_mask_fixupimm_sd): Ditto.
1072         (_mm_mask_fixupimm_ss): Ditto.
1073         * config/i386/avx512vlintrin.h:
1074         (_mm256_fixupimm_pd): Update parameters and builtin.
1075         (_mm256_maskz_fixupimm_pd): Ditto.
1076         (_mm256_fixupimm_ps): Ditto.
1077         (_mm256_maskz_fixupimm_ps): Ditto.
1078         (_mm_fixupimm_pd): Ditto.
1079         (_mm_maskz_fixupimm_pd): Ditto.
1080         (_mm_fixupimm_ps): Ditto.
1081         (_mm_maskz_fixupimm_ps): Ditto.
1082         (_mm256_mask_fixupimm_pd): Update builtin.
1083         (_mm256_mask_fixupimm_ps): Ditto.
1084         (_mm_mask_fixupimm_pd): Ditto.
1085         (_mm_mask_fixupimm_ps): Ditto.
1086         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
1087         * config/i386/i386-builtin.def: Update builtin definitions.
1088         * config/i386/i386.c: Handle new builtin types and remove useless ones.
1089         * config/i386/sse.md: Update VFIXUPIMM* patterns.
1090         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1091         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1092         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
1093         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1094         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1095         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
1096         * config/i386/subst.md:
1097         (round_saeonly_sd_mask_operand4): Add new subst_attr.
1098         (round_saeonly_sd_mask_op4): Ditto.
1099         (round_saeonly_expand_operand5): Ditto.
1100         (round_saeonly_expand): Update.
1102 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
1104         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
1106 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
1108         PR rtl-optimization/87871
1109         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
1111 2018-11-05  Paul Koning  <ni1d@arrl.net>
1113         * doc/sourcebuild.texi (target attributes): Document new "inf"
1114         effective target keyword.
1116 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
1118         * config/s390/s390.c (s390_register_move_cost): Increase costs for
1119         moves involving the CC reg.
1121 2018-11-05  Richard Biener  <rguenther@suse.de>
1123         PR tree-optimization/87873
1124         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
1125         argument.
1126         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
1127         * tree-vect-loop.c (vect_transform_loop): When splitting the
1128         loop exit also create forwarder PHIs for constants.
1129         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
1130         Handle constant to_arg, add extra checking we match up the correct
1131         PHIs.
1133 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
1135         * config/s390/s390.md: QImode and HImode for load on condition.
1137 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
1139         * config/s390/predicates.md: Fix typo.
1140         * config/s390/s390.md: Allow immediates for load on condition.
1142 2018-11-05  Martin Liska  <mliska@suse.cz>
1144         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
1145         * bitmap.h (struct bitmap_usage): Likewise.
1146         * ggc-common.c (SCALE): Remove.
1147         (LABEL): Likewise.
1148         (struct ggc_usage): Use SIZE_AMOUNT. And update
1149         compare method.
1150         * ggc-page.c (SCALE): Remove.
1151         (STAT_LABEL): Remove.
1152         (ggc_print_statistics): Use SIZE_AMOUNT.
1153         * gimple.h (SCALE): Remove.
1154         (LABEL): Likewise.
1155         * input.c (ONE_K): Remove.
1156         (ONE_M): Likewise.
1157         (SCALE): Likewise.
1158         (STAT_LABEL): Likewise.
1159         (FORMAT_AMOUNT): Likewise.
1160         (dump_line_table_statistics): Use SIZE_AMOUNT.
1161         * mem-stats.h (struct mem_usage): Likewise.
1162         * rtl.c (dump_rtx_statistics): Likewise.
1163         (rtx_alloc_counts): Change type to size_t.
1164         (rtx_alloc_sizes): Likewise.
1165         (rtx_count_cmp): New.
1166         (dump_rtx_statistics): Sort first based on counts.
1167         * tree.c (tree_nodes_cmp): New.
1168         (tree_codes_cmp): New.
1169         (dump_tree_statistics): Sort first based on counts.
1170         * system.h (ONE_K): New.
1171         (ONE_M): Likewise.
1172         (SIZE_SCALE): Likewise.
1173         (SIZE_LABEL): Likewise.
1174         (SIZE_AMOUNT): Likewise.
1175         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
1176         * tree-dfa.c (dump_dfa_stats): Likewise.
1177         * tree-phinodes.c (phinodes_print_statistics): Likewise.
1178         * tree-ssanames.c (ssanames_print_statistics): Likewise.
1179         * tree.c (dump_tree_statistics): Likewise.
1180         * vec.c (struct vec_usage): Likewise.
1181         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
1182         trigger a -Werror=format-overflow with
1183         --enable-gather-detailed-stats.
1185 2018-11-05  Martin Liska  <mliska@suse.cz>
1187         * mem-stats.h (mem_alloc_description::release_instance_overhead):
1188         Return T *.
1189         * vec.c (struct vec_usage): Register m_element_size.
1190         (vec_prefix::register_overhead): New arguments: elements and
1191         element_size.
1192         (vec_prefix::release_overhead): Subtract elements.
1193         * vec.h (struct vec_prefix): Change signature.
1194         (va_heap::reserve): Pass proper arguments.
1195         (va_heap::release): Likewise.
1197 2018-11-05  Martin Liska  <mliska@suse.cz>
1199         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
1200         style.
1201         * vec.c: Likewise.
1203 2018-11-05  Richard Biener  <rguenther@suse.de>
1205         * tree-scalar-evolution.h (final_value_replacement_loop): Update
1206         prototype.
1207         * tree-scalar-evolution.c (final_value_replacement_loop): Return
1208         whether anything was done.
1209         (scev_const_prop): Remove constant propagation part, fold
1210         remains into ...
1211         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
1212         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
1213         conditionally.
1215 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
1217         PR tree-optimization/87859
1218         * gimple-ssa-store-merging.c (struct merged_store_group): Add
1219         only_constants and first_nonmergeable_order members.
1220         (merged_store_group::merged_store_group): Initialize them.
1221         (merged_store_group::do_merge): Clear only_constants member if
1222         adding something other than INTEGER_CST store.
1223         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
1224         stores with order >= first_nonmergeable_order.  Use
1225         merged_store->only_constants instead of always recomputing it.
1226         Set merged_store->first_nonmergeable_order if we've skipped any
1227         stores.  Attempt to merge overlapping INTEGER_CST stores that
1228         we would otherwise skip.
1230         PR sanitizer/87837
1231         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
1233 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
1235         PR target/87853
1236         * config/i386/emmintrin.h (__v16qs): New to cope with option
1237         -funsigned-char.
1238         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
1239         (_mm_cmplt_epi8): Likewise.
1240         (_mm_cmpgt_epi8): Likewise.
1242 2018-11-05  Richard Biener  <rguenther@suse.de>
1244         PR rtl-optimization/87852
1245         * fwprop.c (use_killed_between): Only consider single-defs of the
1246         use whose definition statement dominates the use.
1248 2018-11-05  Martin Liska  <mliska@suse.cz>
1250         PR web/87829
1251         * doc/invoke.texi: Remove options that are
1252         not disabled with -Os.
1254 2018-11-05  Martin Liska  <mliska@suse.cz>
1256         PR c/87811
1257         * doc/extend.texi: Update constrain about the last argument
1258         of __builtin_expect_with_probability.
1260 2018-11-05  Martin Liska  <mliska@suse.cz>
1262         PR c/87811
1263         * predict.c (expr_expected_value_1): Verify
1264         that last argument is a real constants and emit
1265         error.
1267 2018-11-05  Martin Liska  <mliska@suse.cz>
1269         PR gcov-profile/77698
1270         * ipa-profile.c (ipa_profile): Adjust hotness threshold
1271         only in LTO mode.
1273 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1275         PR tree-optimization/86572
1276         * builtins.c (c_strlen): Handle negative offsets in a safe way.
1278 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1280         PR tree-optimization/87672
1281         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
1282         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
1284 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
1286         PR middle-end/58372
1287         * cfgexpand.c (pass_expand::execute): Move the call to
1288         finish_eh_generation in front of the call to expand_stack_alignment.
1290 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
1292         * common/config/i386/i386-common.c (processor_alias_table): Add
1293         znver2 entry.
1294         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
1295         (case ${target}): Add znver2.
1296         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
1297         -march=native recognize znver2 processors.
1298         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
1299         * config/i386/i386.c (m_znver2): New definition.
1300         (m_ZNVER): New definition.
1301         (m_AMD_MULTIPLE): Includes m_znver2.
1302         (processor_cost_table): Add znver2 entry.
1303         (processor_target_table): Add znver2 entry.
1304         (get_builtin_code_for_version): Set priority for
1305         PROCESSOR_ZNVER2.
1306         (processor_model): Add M_AMDFAM17H_ZNVER2.
1307         (arch_names_table): Ditto.
1308         (ix86_reassociation_width): Include znver2.
1309         * config/i386/i386.h (TARGET_znver2): New definition.
1310         (struct ix86_size_cost): Add TARGET_ZNVER2.
1311         (enum processor_type): Add PROCESSOR_ZNVER2.
1312         * config/i386/i386.md (define_attr "cpu"): Add znver2.
1313         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
1314         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
1315         (ix86_adjust_cost): Add znver2.
1316         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
1317         * gcc/doc/extend.texi: Add details about znver2.
1318         * gcc/doc/invoke.texi: Add details about znver2.
1320 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
1322         PR target/87079
1324         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
1325         pattern.
1327 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
1329         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
1330         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
1332 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
1334         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
1335         shrink-wrapping when -mtrack-speculation.
1337 2018-11-02  Richard Biener  <rguenther@suse.de>
1339         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
1340         (pop_cost_one_pair): Do not free pair.
1341         (pop_best_coalesce): Likewise.
1342         (create_coalesce_list): Initialize obstack.
1343         (delete_coalesce_list): Free obstack.
1344         (find_coalesce_pair): Obstack-allocate coalesce pairs.
1345         (add_cost_one_coalesce): Likewise.
1346         (struct live_track): Remove bitmap pointer indirections.
1347         (new_live_track): Adjust.
1348         (delete_live_track): Likewise.
1349         (live_track_remove_partition): Likewise.
1350         (live_track_add_partition): Likewise.
1351         (live_track_live_p): Likewise.
1352         (live_track_process_def): Likewise.
1353         (live_track_clear_base_vars): Likewise.
1355 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1357         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
1358         * configure: Regenerate.
1359         * config.in: Regenerate.
1360         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
1361         HAVE_AS_REGISTER_PSEUDO_OP guard.
1362         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
1363         !HAVE_AS_REGISTER_PSEUDO_OP support.
1364         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
1365         guard.
1367 2018-11-02  Richard Biener  <rguenther@suse.de>
1369         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
1370         with is_type_die.
1372 2018-11-02  Richard Biener  <rguenther@suse.de>
1374         PR tree-optimization/87776
1375         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
1376         executable when iterating but running into rpo-vn-max-loop-depth
1377         and not eliding the iteration.
1379 2018-11-30  Jan Hubicka  <jh@suse.cz>
1381         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
1382         TYPE_DECL.
1384 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
1386         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
1387         prototype.
1388         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
1389         Combine with rs6000_address_for_fpconvert.
1390         (rs6000_address_for_fpconvert): Combine with
1391         rs6000_force_indexed_or_indirect_mem.
1392         (rs6000_expand_vector_init): Change function call from
1393         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
1394         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
1395         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
1396         (floatsi<mode>2_lfiwax_mem): Ditto.
1397         (floatunssi<mode>2_lfiwzx): Ditto.
1398         (floatunssi<mode>2_lfiwzx_mem): Ditto.
1399         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
1400         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
1401         (fix_trunc<mode>si2_stfiwx): Ditto.
1402         (fixuns_trunc<mode>si2_stfiwx): Ditto.
1403         (float_<mode>si2_hw): Ditto.
1404         (floatuns_<mode>si2_hw): Ditto.
1405         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
1406         (vsx_splat_<mode>): Ditto.
1408 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
1410         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
1411         at end of assembler input text.
1412         * configure: Regenerate.
1414 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
1416         PR tree-optimization/87826
1417         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
1418         negative or larger or equal to type's precision.
1420 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
1422         * opts.c (default_options_table): Do not enable
1423         OPT_fdelayed_branch at -Og.
1424         * doc/invoke.texi (-fdelayed-branch): Document it.
1426 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
1428         * optabs-libfuncs.c (build_libfunc_function_visibility):
1429         New, split out from...
1430         (build_libfunc_function): ... here.
1431         (init_one_libfunc_visibility): New, split out from ...
1432         (init_one_libfunc): ... here.
1434         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
1435         scratch register need not be early-clobber.  Document the reason
1436         why we cannot use ST<OP>.
1438 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
1440         PR bootstrap/82856
1441         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
1442         line for second argument of AC_DEFINE_UNQUOTED.
1443         * doc/install.texi (Tools/packages necessary for modifying GCC):
1444         Update to autoconf 2.69 and automake 1.15.1.
1445         * aclocal.m4, config.in, configure: Regenerate.
1447 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
1449         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
1450         initialization.
1452 2018-10-31  Martin Liska  <mliska@suse.cz>
1454         PR driver/83193
1455         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
1456         Add new argument invalid_extension.
1457         (aarch64_get_all_extension_candidates): New function.
1458         (aarch64_rewrite_selected_cpu): Add NULL to function call.
1459         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
1460         new argument.
1461         (aarch64_get_all_extension_candidates): New function.
1462         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
1463         argument invalid_extension.
1464         (aarch64_parse_cpu): Likewise.
1465         (aarch64_print_hint_for_extensions): New function.
1466         (aarch64_validate_mcpu): Provide hint about invalid extension.
1467         (aarch64_validate_march): Likewise.
1468         (aarch64_handle_attr_arch): Pass new argument.
1469         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
1470         (aarch64_handle_attr_isa_flags): Likewise.
1472 2018-10-31  Richard Biener  <rguenther@suse.de>
1474         PR middle-end/70359
1475         PR middle-end/86270
1476         * tree-outof-ssa.c (insert_backedge_copies): Restrict
1477         copy generation to useful cases.  Place the copy before
1478         the definition of the backedge value when possible.
1480 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
1482         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
1483         * config/arc/arc.c (arc_active_insn): New function.
1484         (check_store_cacheline_hazard): Likewise.
1485         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
1486         (arc_override_options): Disable delay slot scheduler for older
1487         A7.
1488         (arc_store_addr_hazard_p): New implementation, old one renamed to
1489         ...
1490         (arc_store_addr_hazard_internal_p): Renamed.
1491         (arc_reorg): Don't combine into brcc instructions which are part
1492         of hardware hazard solution.
1493         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
1494         (tune_arc700): Likewise.
1495         * config/arc/arc.opt (arc7xx): New tune value.
1496         * config/arc/arc700.md: Improve A7 scheduler.
1498 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
1500         * config/arc/arc.c (arc_override_options): Remove
1501         TARGET_COMPACT_CASESI.
1502         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
1503         (CASE_VECTOR_MODE): Likewise.
1504         (CASE_VECTOR_PC_RELATIVE): Likewise.
1505         (CASE_VECTOR_SHORTEN_MODE): Likewise.
1506         (CASE_VECTOR_SHORTEN_MODE1): Delete.
1507         (ADDR_VEC_ALIGN): Update.
1508         (ASM_OUTPUT_CASE_LABEL): Undefine.
1509         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
1510         (TARGET_BI_BIH): Define.
1511         (DEFAULT_BRANCH_INDEX): Likewise.
1512         * config/arc/arc.md (casesi): Rework to accept BI/BIH
1513         instructions, remove compact_casesi use case.
1514         (casesi_compact_jump): Remove.
1515         (casesi_dispatch): New pattern.
1516         * config/arc/arc.opt: Add mbranch-index option. Deprecate
1517         compact_casesi option.
1518         * doc/invoke.texi: Document mbranch-index option.
1520 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
1522         * config/arc/arc.c (arc_get_tp): Remove function.
1523         (arc_emit_call_tls_get_addr): Likewise.
1524         (arc_call_tls_get_addr): New function.
1525         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
1526         * config/arc/arc.md (tls_load_tp_soft): Remove.
1527         (tls_gd_get_addr): Likewise.
1529 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
1531         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
1532         (mulsi3_600_lib): Remove pattern.
1533         (umulsi3_highpart_600_lib_le): Likewise.
1534         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
1535         (umulsidi3): Remove call to umulsidi3_600_lib.
1536         (umulsidi3_600_lib): Remove pattern.
1537         (peephole2): Remove peephole using the above deprecated patterns.
1539 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
1541         PR target/87374
1542         * config/arm/arm.c (arm_option_check_internal): Disable the combined
1543         use of -mslow-flash-data and -mword-relocations.
1544         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
1545         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
1546         flag_pic.
1547         * doc/invoke.texi (-mword-relocations): Mention conflict with
1548         -mslow-flash-data.
1549         (-mslow-flash-data): Reciprocally.
1551 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
1553         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
1554         16-byte modes held in GP registers to use an even regno.
1556         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
1557         (aarch64_atomic_ldop_supported_p): Remove.
1558         (aarch64_gen_atomic_ldop): Remove.
1559         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
1560         Fully expand LSE operations here.
1561         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
1562         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
1563         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
1564         and use ATOMIC_LDOP instead; use register_operand for the input;
1565         drop the split and emit insns directly.
1566         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
1567         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
1568         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
1570         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
1571         (aarch64_gen_atomic_ldop): Don't call it.
1572         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
1573         Use aarch64_reg_or_zero.
1574         (aarch64_atomic_exchange<ALLI>): Likewise.
1575         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
1576         operand 0; use aarch64_reg_or_zero for input; merge ...
1577         (@aarch64_atomic_swp<ALLI>): ... this and remove.
1579         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
1580         (aarch64_split_compare_and_swap): Use it.
1581         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
1582         test oldval against the proper predicate.
1583         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
1584         Use nonmemory_operand for expected.
1585         (cas_short_expected_pred): New.
1586         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
1587         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
1588         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
1589         (aarch64_plushi_operand): New.
1591         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
1592         Force oldval into the rval register for TARGET_LSE; emit the compare
1593         during initial expansion so that it may be deleted if unused.
1594         (aarch64_gen_atomic_cas): Remove.
1595         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
1596         Change =&r to +r for operand 0; use match_dup for operand 2;
1597         remove is_weak and mod_f operands as unused.  Drop the split
1598         and merge with...
1599         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
1600         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
1601         (@aarch64_atomic_cas<GPI>): Similarly.
1603 2018-10-31  Richard Biener  <rguenther@suse.de>
1605         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
1606         using ABSU_EXPR.
1608 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1610         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
1611         * config/aarch64/aarch64.md: Include saphira.md
1612         * config/aarch64/saphira.md: New file for pipeline description.
1614 2018-10-30  Martin Sebor  <msebor@redhat.com>
1616         PR middle-end/87041
1617         * gimple-ssa-sprintf.c (format_directive): Use %G to include
1618         inlining context.
1619         (sprintf_dom_walker::compute_format_length):
1620         Avoid setting POSUNDER4K here.
1621         (get_destination_size): Handle null argument values.
1622         (get_user_idx_format): New function.
1623         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
1624         functions, including user-defined with attribute format printf.
1625         Use %G to include inlining context.
1626         Set POSUNDER4K here.
1628 2018-10-30  Jan Hubicka  <jh@suse.cz>
1630         * params.def (lto-partitions): Bump from 32 to 128.
1632 2018-10-30  Jan Hubicka  <jh@suse.cz>
1634         * tree.c
1635         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
1636         head in file.
1637         (free_lang_data_in_type): Forward declare.
1638         (fld_type_variant_equal_p): New function.
1639         (fld_type_variant): New function
1640         (fld_incomplete_types): New hash.
1641         (fld_incomplete_type_of): New function
1642         (fld_simplfied-type): New function.
1643         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
1644         (free_lang_data): Allocate and free fld_incomplete_type; update call
1645         of free_lang_data_in_decl.
1647 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
1649         * gcov.c (output_lines): Remove duplicate line.
1651 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
1653         * config/rs6000/rs6000.md (bswapdi2): Force address into register
1654         if not in indexed or indirect form.
1655         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
1656         (bswapdi2_store): Ditto.
1657         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
1658         helper function.
1659         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
1660         Prototype for helper function.
1662 2018-10-30  Martin Sebor  <msebor@redhat.com>
1664         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
1665         (target, pragma GCC optimize, pragma GCC target): Ditto.
1667 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
1669         * doc/extend.texi: Fix prototype and description of
1670         __builtin_expect_with_probability.
1672 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
1674         * cgraph.h (clone_function_name_1): Replaced by new
1675           clone_function_name_numbered that takes name as string; for
1676           privatize_symbol_name_1 use only.
1677           (clone_function_name): Renamed to
1678           clone_function_name_numbered to be explicit about numbering.
1679           (clone_function_name): New two-argument function that does
1680           not number its output.
1681           (clone_function_name): New three-argument function that
1682           takes a number to append to its output.
1683         * cgraphclones.c (duplicate_thunk_for_node):
1684           (clone_function_name_1): Renamed.
1685           (clone_function_name_numbered): Two new functions.
1686           (clone_function_name): Improved documentation.
1687           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
1688         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
1689         * final.c (final_scan_insn_1): Use the new clone_function_name
1690           without numbering.
1691         * multiple_target.c (create_dispatcher_calls): Ditto.
1692           (create_target_clone): Ditto.
1693         * omp-expand.c (grid_expand_target_grid_body): Ditto.
1694         * omp-low.c (create_omp_child_function_name): Ditto.
1695         * omp-simd-clone.c (simd_clone_create): Ditto.
1696         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
1697           new clone_function_name without numbering.
1699 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
1701         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
1702         Assert that the allocation size is not zero.
1704 2018-10-30  Richard Biener  <rguenther@suse.de>
1706         PR tree-optimization/87800
1707         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
1708         non-induction or reduction PHIs.
1710 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
1712         * config/aarch64/falkor-tag-collision-avoidance.c
1713         (execute_tag_collision_avoidance): Call df_note_add_problem.
1715 2018-10-30  Martin Liska  <mliska@suse.cz>
1717         * doc/extend.texi: Fix typo in documentation
1718         of __builtin_expect_with_probability.
1720 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
1722         PR c++/87721
1723         * input.c (get_substring_ranges_for_loc): Detect if
1724         linemap_resolve_location gives us a NULL map, and reject
1725         this case.
1727 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
1729         * config.gcc (xstormy16-*-elf): Set tm_d_file.
1731 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
1732             Martin Sebor  <msebor@redhat.com>
1733             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
1735         * doc/ux.texi (Quoting): New subsection, adapted from material at
1736         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
1737         MartinSebor and ManuelLopezIbanez.
1738         (Fix-it hints): Note that fix-it hints shouldn't be marked for
1739         translation.
1741 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
1743         PR middle-end/87469
1744         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
1745         max value.
1747 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
1749         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
1751 2018-10-29  Paul Koning  <ni1d@arrl.net>
1753         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
1755 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
1757         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
1758         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
1759         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
1760         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
1761         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
1762         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
1763         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
1764         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
1765         Change 'vector' to '__vector'.
1766         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
1767         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
1768         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
1769         _mm_avg_pu16): Likewise.  And, whitespace corrections.
1771 2018-10-29  Richard Biener  <rguenther@suse.de>
1773         PR tree-optimization/87785
1774         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
1775         internal defs.
1777 2018-10-29  Olivier Hainque  <hainque@adacore.com>
1779         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
1781 2018-10-29  Olivier Hainque  <hainque@adacore.com>
1783         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
1784         ports configuration macro, defaults to "gnu".
1785         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
1786         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
1788 2018-10-29  Olivier Hainque  <hainque@adacore.com>
1790         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
1792 2018-10-29  Richard Biener  <rguenther@suse.de>
1794         PR tree-optimization/87790
1795         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
1796         (vect_make_slp_decision): Adjust.
1797         (vect_slp_analyze_bb_1): Likewise.
1798         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
1799         edges.
1801 2018-10-29  Richard Biener  <rguenther@suse.de>
1803         PR tree-optimization/87785
1804         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
1805         and processing.
1806         (vect_build_slp_tree): Likewise.
1807         (vect_gather_slp_loads): New function.
1808         (vect_analyze_slp_instance): Gather loads separately from the
1809         SLP tree build.
1811 2018-10-29  Martin Liska  <mliska@suse.cz>
1813         * Makefile.in: Make dependency to json.o.
1814         * doc/gcov.texi: Document new JSON format, remove
1815         old intermediate format documentation.
1816         * gcov.c (struct function_info): Come up with m_name and
1817         m_demangled_name.
1818         (function_info::function_info): Initialize it.
1819         (function_info::~function_info): Release it.
1820         (main): Rename flag_intermediate_format to flag_json_format.
1821         (print_usage): Describe --json-format.
1822         (process_args): Set flag_json_format.
1823         (output_intermediate_line): Remove.
1824         (output_intermediate_json_line): Likewise.
1825         (get_gcov_intermediate_filename): Return new extension
1826         ".gcov.json.gz".
1827         (output_intermediate_file): Implement JSON emission.
1828         (output_json_intermediate_file): Implement JSON emission.
1829         (generate_results): Use ::get_name for function name.
1830         Handle JSON output file.
1831         (read_graph_file): Use ::get_name instead of cplus_demangle.
1832         (read_count_file): Likewise.
1833         (solve_flow_graph): Likewise.
1834         (add_line_counts): Likewise.
1835         (accumulate_line_counts): Use new flag_json_format.
1836         (output_function_details): Use ::get_name instead of cplus_demangle.
1837         (output_lines): Likewise.
1838         * json.cc (test_writing_literals): Add new tests.
1839         * json.h (class literal): Add new boolean constructor.
1841 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
1843         PR rtl-optimization/87701
1844         PR rtl-optimization/87780
1845         * combine.c (make_more_copies): Rewrite.
1847 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
1849         * doc/generic.texi (ABSU_EXPR): Document.
1850         * match.pd (absu(x)*absu(x) -> x*x): Handle.
1851         (absu(absu(X)) -> absu(X)): Likewise.
1852         (absu(-X) -> absu(X)): Likewise.
1853         (absu(X)  where X is nonnegative -> X): Likewise.
1855 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
1857         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
1858         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
1859         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
1860         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
1861         (s-tm-texi): Also check timestamp on d-target.def.
1862         (generated_files): Add TM_D_H and d-target-hooks-def.h.
1863         (build/genhooks.o): Also depend on D_TARGET_DEF.
1864         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
1865         variables.
1866         * config/aarch64/aarch64-d.c: New file.
1867         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
1868         Define.
1869         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
1870         prototype.
1871         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
1872         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
1873         * config/arm/arm-d.c: New file.
1874         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
1875         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
1876         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
1877         * config/arm/t-arm (arm-d.o): New rule.
1878         * config/default-d.c: New file.
1879         * config/glibc-d.c: New file.
1880         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1881         * config/i386/i386-d.c: New file.
1882         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
1883         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
1884         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
1885         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
1886         * config/i386/t-i386 (i386-d.o): New rule.
1887         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1888         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1889         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
1890         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1891         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
1892         * config/mips/mips-d.c: New file.
1893         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
1894         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
1895         * config/mips/t-mips (mips-d.o): New rule.
1896         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1897         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1898         * config/powerpcspe/powerpcspe-d.c: New file.
1899         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
1900         New prototype.
1901         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
1902         Support GNU D by using 0 as the language type.
1903         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
1904         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
1905         * config/riscv/riscv-d.c: New file.
1906         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
1907         prototype.
1908         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
1909         * config/riscv/t-riscv (riscv-d.o): New rule.
1910         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1911         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
1912         * config/rs6000/rs6000-d.c: New file.
1913         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
1914         prototype.
1915         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
1916         Support GNU D by using 0 as the language type.
1917         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
1918         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
1919         * config/s390/s390-d.c: New file.
1920         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
1921         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
1922         * config/s390/t-s390 (s390-d.o): New rule.
1923         * config/sparc/sparc-d.c: New file.
1924         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
1925         prototype.
1926         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
1927         * config/sparc/t-sparc (sparc-d.o): New rule.
1928         * config/t-glibc (glibc-d.o): New rule.
1929         * configure: Regenerated.
1930         * configure.ac (tm_d_file): New variable.
1931         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
1932         * doc/contrib.texi (Contributors): Add self for the D frontend.
1933         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
1934         * doc/install.texi (Configuration): Mention libphobos as an option for
1935         --enable-shared.  Mention d as an option for --enable-languages.
1936         (Testing): Mention check-d as a target.
1937         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
1938         name suffixes.  Mention d as a -x option.
1939         * doc/sourcebuild.texi (Top Level): Mention libphobos.
1940         * doc/standards.texi (Standards): Add section on D language.
1941         * doc/tm.texi: Regenerated.
1942         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
1943         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
1944         * dwarf2out.c (is_dlang): New function.
1945         (gen_compile_unit_die): Use DW_LANG_D for D.
1946         (declare_in_namespace): Return module die for D, instead of adding
1947         extra declarations into the namespace.
1948         (gen_namespace_die): Generate DW_TAG_module for D.
1949         (gen_decl_die): Handle CONST_DECLSs for D.
1950         (dwarf2out_decl): Likewise.
1951         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
1952         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
1953         kinds of aggregates.
1954         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
1955         * genhooks.c: Include d/d-target.def.
1957 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
1959         PR target/85669
1960         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
1961         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
1962         (STACK_DYNAMIC_OFFSET): Likewise.
1964 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
1966         PR target/80024
1967         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
1968         error message.
1970 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
1972         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
1973         return type and other typos.
1975 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
1977         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
1978         a shorter sequence with fewer branches.
1979         (emit_final_str_compare_gpr): Ditto.
1981 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
1983         * config/rs6000/tmmintrin.h: New file.
1984         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
1986 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
1988         * config/rs6000/mmintrin.h: Enable 32bit compilation.
1989         * config/rs6000/xmmintrin.h: Likewise.
1991 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
1993         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
1995 2018-10-26  Richard Biener  <rguenther@suse.de>
1997         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
1998         and wrapper.
1999         (vect_mark_slp_stmts_relevant): Likewise.
2000         (vect_detect_hybrid_slp_stmts): Likewise.
2001         (vect_bb_slp_scalar_cost): Likewise.
2002         (vect_remove_slp_scalar_calls): Likewise.
2004 2018-10-26  Jan Hubicka  <jh@suse.cz>
2006         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
2007         (warn_types_mismatch): Fix walk of DECL_NAME.
2008         (odr_types_equivalent_p): Fix overactive assert.
2010 2018-10-26  Richard Biener  <rguenther@suse.de>
2012         PR tree-optimization/87105
2013         * tree-vectorizer.h (_slp_tree::refcnt): New member.
2014         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
2015         refcnt.
2016         (vect_create_new_slp_node): Initialize refcnt to one.
2017         (bst_traits): Move.
2018         (scalar_stmts_set_t, bst_fail): Remove.
2019         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
2020         (vect_build_slp_tree): Add bst_map argument and lookup
2021         already created SLP nodes.
2022         (vect_print_slp_tree): Handle a SLP graph, print SLP node
2023         addresses.
2024         (vect_slp_rearrange_stmts): Handle a SLP graph.
2025         (vect_analyze_slp_instance): Adjust and free SLP nodes from
2026         the CSE map.  Fix indenting.
2027         (vect_schedule_slp_instance): Add short-cut.
2029 2018-10-26  Martin Liska  <mliska@suse.cz>
2031         PR testsuite/86158
2032         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
2033         addr_expr and not with pointers.
2035 2018-10-26  Jan Hubicka  <jh@suse.cz>
2037         * tree.c (free_lang_data_in_type): Only check main variants.
2038         * ipa-devirt.c (warn_odr): Make static.
2039         (types_same_for_odr): Drop strict variant.
2040         (types_odr_comparable): Likewise.
2041         (odr_or_derived_type_p): Look for main variants.
2042         (odr_name_hasher::equal): Cleanup comment.
2043         (odr_subtypes_equivalent): Add warn and warned arguments; check main
2044         variants.
2045         (type_variants_equivalent_p): break out from ...
2046         (odr_types_equivalent): ... here; go for main variants where needed.
2047         (warn_odr): ... here; turn static.
2048         (warn_types_mismatch): Compare mangled names of main variants.
2049         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
2050         (type_with_linkage_p): Sanity check that we look at main variant.
2051         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
2052         * tree.h (types_same_for_odr): Drop strict argument.
2054 2018-10-26  Richard Biener  <rguenther@suse.de>
2056         PR tree-optimization/87746
2057         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
2058         Simplify and fix WRT strided store groups with size not
2059         equal to step in element count.
2060         (vect_analyze_group_access_1): Dump the whole group.
2062 2018-10-25  Carl Love  <cel@us.ibm.com>
2064         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
2065         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
2066         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
2067         precicion entry for each overloaded builtin.
2068         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
2069         VSCEDPUO): Rename overloaded name.
2070         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
2071         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
2072         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
2073         define_expand for xscmpexqp instruction.
2074         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
2076 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2077             Jinsong Ji <jji@us.ibm.com>
2079         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
2080         function with vec_sl.
2081         (_mm_slli_epi32): Likewise.
2082         (_mm_slli_epi64): Likewise.
2083         (_mm_srai_epi16): Replace deprecated function with vec_sra.
2084         (_mm_srai_epi32): Likewise.
2085         (_mm_srli_epi16): Replace deprecated function with vec_sr.
2086         (_mm_srli_epi32): Likewise.
2087         (_mm_srli_epi64): Likewise.
2088         (_mm_sll_epi16): Replace deprecated function with vec_sl.
2089         (_mm_sll_epi32): Likewise.
2090         (_mm_sll_epi64): Likewise.
2091         (_mm_sra_epi16): Replace deprecated function with vec_sra.
2092         (_mm_sra_epi32): Likewise.
2093         (_mm_srl_epi16): Replace deprecated function with vec_sr.
2094         (_mm_srl_epi32): Likewise.
2095         (_mm_srl_epi64): Likewise.
2097 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2098             Jinsong Ji <jji@us.ibm.com>
2100         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
2101         comparison operators with vec_cmp* for compatibility due to
2102         unfortunate history; clean up formatting and use types more
2103         appropriately.
2104         (_mm_sll_epi32): Likewise.
2105         (_mm_sll_epi64): Likewise.
2106         (_mm_srl_epi16): Likewise.
2107         (_mm_srl_epi32): Likewise.
2108         (_mm_srl_epi64): Likewise.
2110 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
2111             Jinsong Ji <jji@us.ibm.com>
2113         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
2114         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
2115         __vector __bool int.  Use vec_cmpgt in preference to deprecated
2116         function vec_vcmpgtfp.
2117         (_mm_max_ps): Likewise.
2119 2018-10-25  Jeff Law  <law@redhat.com>
2121         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
2122         if returning false.
2124 2018-10-25  Martin Sebor  <msebor@redhat.com>
2126         * doc/extend.texi (aligned): Expand attribute description.
2127         (Alignment): Rename section.  Discuss function arguments.
2129 2018-10-25  Jan Hubicka  <jh@suse.cz>
2131         * ipa-devirt.c (main_odr_variant): Remove.
2132         (hash_odr_name, types_same_for_odr, types_odr_comparable,
2133         odr_name_hasher::equal, odr_subtypes_equivalent_p):
2134         Drop use of main_odr_variant.
2135         (add_type_duplicate): Silence confused warnings on integer types.
2136         (get_odr_type): Always look for main variant.
2137         (register_odr_type): Simplify.
2139 2018-10-25  Richard Biener  <rguenther@suse.de>
2141         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
2142         Initialize ng to silence error with release checking bootstrap.
2144 2018-10-25  Richard Biener  <rguenther@suse.de>
2146         * tree-if-conv.c: Include tree-ssa-sccvn.h.
2147         (tree_if_conversion): Run CSE on the if-converted loop body.
2149 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
2151         * config/s390/constraints.md (ZL): New constraint.
2152         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
2153         operands.
2154         * config/s390/s390.md (movdi_larl): Remove.
2155         (movdi_64): Add the LARL alternative.
2157 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
2159         PR bootstrap/87747
2160         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
2161         (RTX_CODE_HWINT_P): New macro.
2162         (rtx_code_size): Use RTX_CODE_HWINT_P ().
2164 2018-10-25  Jan Hubicka  <jh@suse.cz>
2166         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
2167         is anonymous.
2169 2018-10-25  Richard Biener  <rguenther@suse.de>
2171         PR tree-optimization/87665
2172         PR tree-optimization/87745
2173         * tree-vectorizer.h (get_earlier_stmt): Remove.
2174         (get_later_stmt): Pick up UID from the original non-pattern stmt.
2176 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
2178         * options.texi (Deprecated): Move list to Var section.
2180 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
2181             Jinsong Ji <jji@us.ibm.com>
2183         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
2184         __vector long to __vector long long.
2185         (_mm_cvtpd_ps): Likewise.
2186         (_mm_cvttpd_epi32): Likewise.
2187         (_mm_cvtpi32_pd): Likewise.
2188         (_mm_unpackhi_epi64): Likewise.
2189         (_mm_unpacklo_epi64): Likewise.
2191 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
2193         PR rtl-optimization/87720
2194         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
2196 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
2198         * gimple-ssa-isolate-paths.c
2199         (find_implicit_erroneous_behavior): Do not change code if the
2200         pass is running for warnings only.
2201         (find_explicit_erroneous_behavior): Likewise.
2203 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
2205         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
2206         Define as rs6000_mangle_decl_assembler_name.
2207         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
2208         long double to IEEE long double, switch the names of the long
2209         double built-in functions to be <func>f128 instead of <func>l.
2211 2018-10-24  Martin Sebor  <msebor@redhat.com>
2213         * doc/extend.texi (nonnull): List no-argument form.  Reference
2214         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
2216 2018-10-24  Richard Biener  <rguenther@suse.de>
2218         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
2220 2018-10-24  Martin Liska  <mliska@suse.cz>
2222         PR tree-optimization/84436
2223         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
2224         Remove.
2225         (switch_conversion::contains_linear_function_p): New.
2226         (switch_conversion::build_one_array): Support linear
2227         transformation on input.
2228         * tree-switch-conversion.h (struct switch_conversion): Add
2229         contains_linear_function_p declaration.
2231 2018-10-24  Richard Biener  <rguenther@suse.de>
2233         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
2234         if its argument is CONSTANT_CLASS_P.
2236 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
2238         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
2239         it is wrong for forward declarations.
2241 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
2243         * config/s390/s390.c (s390_check_qrst_address): Add the missing
2244         SYMBOL_REF_P () check.
2246 2018-10-24  Richard Biener  <rguenther@suse.de>
2248         PR tree-optimization/87105
2249         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
2250         dump classification.
2251         (vect_analyze_data_ref_accesses): Handle duplicate loads and
2252         stores by splitting the affected group after the fact.
2253         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
2254         fail the SLP build because of size constraints.
2256 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2258         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
2259         * configure: Regenerate.
2260         * config.in: Regenerate.
2261         * varasm.c (mergeable_string_section): Use readonly_data_section
2262         if linker doesn't support SHF_MERGE with alignment > 8.
2263         (mergeable_constant_section): Likewise.
2265 2018-10-24  Richard Biener  <rguenther@suse.de>
2267         PR tree-optimization/84013
2268         * tree-ssa-structalias.c (struct msdi_data): New struct for
2269         marshalling data to walk_stmt_load_store_ops.
2270         (maybe_set_dependence_info): Refactor as callback for
2271         walk_stmt_load_store_ops.
2272         (compute_dependence_clique): Set restrict info on all stmt kinds.
2274 2018-10-24  Martin Liska  <mliska@suse.cz>
2276         * cgraph.c (cgraph_node::dump):
2277         Remove reduntant dumps and make tp_first_run dump more compact.
2279 2018-10-24  Richard Biener  <rguenther@suse.de>
2281         PR tree-optimization/87665
2282         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
2283         to reflect reality.
2285 2018-10-12  Jeff Law  <law@redhat.com>
2287         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
2288         for H8/S.
2290 2018-10-23  Richard Biener  <rguenther@suse.de>
2292         * tree-vrp.c (add_assert_info): Guard dump_printf with
2293         dump_enabled_p.
2294         * gimple-ssa-evrp-analyze.c
2295         (evrp_range_analyzer::record_ranges_from_incoming_edge):
2296         Use value_range::ignore_equivs_equal_p.
2298 2018-10-23  Richard Biener  <rguenther@suse.de>
2300         PR tree-optimization/87105
2301         PR tree-optimization/87608
2302         * passes.def (pass_all_early_optimizations): Add early phi-opt
2303         after dce.
2304         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
2305         addition to debug stmts.
2306         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
2307         and abs replacement early.
2308         * tree-cfg.c (gimple_empty_block_p): Likewise.
2310 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
2312         PR target/86383
2313         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
2314         specified to configure.
2315         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
2317 2018-10-23  Richard Biener  <rguenther@suse.de>
2319         PR tree-optimization/87700
2320         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
2322 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
2324         PR target/87674
2325         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
2326         second argument from __mmask16 to __mmask8.
2327         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
2328         _mm_mask_packs_epi32): Likewise.
2329         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
2330         Likewise.
2331         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
2333 2018-10-23  Richard Biener  <rguenther@suse.de>
2335         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
2337 2018-10-23  Richard Biener  <rguenther@suse.de>
2339         PR tree-optimization/86144
2340         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
2341         over simd attribute.
2343 2018-10-23  Richard Biener  <rguenther@suse.de>
2345         PR tree-optimization/87693
2346         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
2347         the case we do not find the taken edge.
2349 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
2350             Jinsong Ji  <jji@us.ibm.com>
2352         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
2353         (_mm_store_pd): Use unaligned vector type for pointer cast.
2354         (_mm_maskmoveu_si128): Likewise.
2355         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
2356         (_mm_store_ps): Use unaligned vector type for pointer cast.
2358 2018-10-22  Paul Koning  <ni1d@arrl.net>
2360         * symtab.c (symtab_node::increase_alignment): Correct max
2361         alignment check.
2363 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
2365         PR tree-optimization/87633
2366         * match.pd: Do not generate unordered integer comparisons.
2368 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
2370         PR rtl-optimization/87600
2371         * combine.c: Add include of expr.h.
2372         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
2373         register to a pseudo.
2374         (make_more_copies): New function, add a copy to a new pseudo after
2375         the moves from hard registers into pseudos.
2376         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
2377         later.  Call make_more_copies.
2379 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
2381         * lra-constraints.c (process_alt_operands): New local array,
2382         matching_early_clobber.  Check matching_early_clobber before
2383         decrementing reject, and set matching_early_clobber after.
2385 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
2387         PR target/87598
2388         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
2389         call output_addr_const and hope for the best.
2391 2018-10-22  Richard Biener  <rguenther@suse.de>
2393         * gimple-ssa-evrp-analyze.c
2394         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
2395         smarter about what ranges to use.
2396         * tree-vrp.c (add_assert_info): Dump here.
2397         (register_edge_assert_for_2): Instead of here at multiple but
2398         not all places.
2400         * gcc.dg/tree-ssa/evrp12.c: New testcase.
2401         * gcc.dg/predict-6.c: Adjust.
2402         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
2403         * gcc.dg/tree-ssa/vrp02.c: Likewise.
2404         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
2406 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
2407         Richard Biener  <rguenther@suse.de>
2409         * bitmap.h: Update data structure documentation, including a
2410         description of bitmap views as either linked-lists or splay trees.
2411         (struct bitmap_element_def): Update comments for splay tree bitmaps.
2412         (struct bitmap_head_def): Likewise.
2413         (bitmap_list_view, bitmap_tree_view): New prototypes.
2414         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
2415         tree_form fields.
2416         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
2417         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
2418         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
2419         released bitmap element here.
2420         (bitmap_element_free): Remove.
2421         (bitmap_elt_clear_from): Work on splay tree bitmaps.
2422         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
2423         this function similar ones such that linked-list bitmap implementation
2424         functions are grouped.
2425         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
2426         and moved for grouping.
2427         (bitmap_list_insert_element_after): Renamed from
2428         bitmap_elt_insert_after, and moved for grouping.
2429         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
2430         (bitmap_tree_link_left, bitmap_tree_link_right,
2431         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
2432         bitmap_tree_link_element, bitmap_tree_unlink_element,
2433         bitmap_tree_find_element): New functions for splay-tree bitmap
2434         implementation.
2435         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
2436         Renamed and moved, see above entries.
2437         (bitmap_tree_listify_from): New function to convert part of a splay
2438         tree bitmap to a linked-list bitmap.
2439         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
2440         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
2441         (bitmap_find_bit): Remove.
2442         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
2443         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
2444         Handle splay tree bitmaps.
2445         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
2446         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
2447         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
2448         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
2449         bitmap_intersect_compl_p, bitmap_ior_and_compl,
2450         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
2451         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
2452         corresponding changes to use linked-list specific bitmap_element
2453         manipulation functions as applicable for efficiency.
2454         (bitmap_tree_to_vec): New function.
2455         (debug_bitmap_elt_file): New function split out from ...
2456         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
2457         (bitmap_print): Likewise.
2459         PR tree-optimization/63155
2460         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
2461         SSA edge worklists.
2462         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
2463         in tree-view.
2465 2018-10-22  Martin Liska  <mliska@suse.cz>
2467         PR tree-optimization/87686
2468         Revert
2469         2018-08-29  Martin Liska  <mliska@suse.cz>
2471         * tree-switch-conversion.c (switch_conversion::expand):
2472         Strenghten assumption about gswitch statements.
2474 2018-10-22  Martin Liska  <mliska@suse.cz>
2476         * ipa-icf.c (sem_item::compare_attributes): Remove.
2477         (sem_item::compare_referenced_symbol_properties): Use
2478         attribute_list_equal instead.
2479         (sem_function::equals_wpa): Likewise.
2480         * ipa-icf.h: Remove compare_attributes.
2482 2018-10-22  Richard Biener  <rguenther@suse.de>
2484         PR middle-end/87682
2485         * mem-stats.h (mem_usage::operator==): Fix pasto.
2487 2018-10-22  Richard Biener  <rguenther@suse.de>
2489         PR tree-optimization/87640
2490         * tree-vrp.c (set_value_range_with_overflow): Decompose
2491         incomplete result.
2492         (extract_range_from_binary_expr_1): Adjust.
2494 2018-10-22  Martin Jambor  <mjambor@suse.cz>
2496         * tree-eh.h (stmt_could_throw_p): Add function parameter.
2497         (stmt_can_throw_external): Likewise.
2498         (stmt_can_throw_internal): Likewise.
2499         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
2500         (lower_eh_constructs_2): Likewise.
2501         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
2502         (stmt_can_throw_external): Likewise.
2503         (stmt_can_throw_internal): Likewise.
2504         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
2505         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
2506         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
2507         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
2508         (pass_lower_eh_dispatch::execute): Pass cfun to
2509         stmt_can_throw_external.
2510         (cleanup_empty_eh): Likewise.
2511         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
2512         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
2513         stmt_can_throw_external instead of pushing it to cfun.
2514         (symbol_table::create_edge): Likewise.
2515         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
2516         stmt_can_throw_internal.
2517         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
2518         to stmt_could_throw_p.
2519         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
2520         stmt_can_throw_internal.
2521         (pass_store_merging::execute): Likewise.
2522         * gimple-ssa-strength-reduction.c
2523         (find_candidates_dom_walker::before_dom_children): Pass cfun to
2524         stmt_could_throw_p.
2525         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
2526         stmt_can_throw_internal.
2527         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
2528         to stmt_can_throw_external.
2529         (check_stmt): Pass cfun to stmt_could_throw_p.
2530         (check_stmt): Pass cfun to stmt_can_throw_external.
2531         (pass_nothrow::execute): Likewise.
2532         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
2533         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
2534         stmt_can_throw_internal.
2535         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
2536         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
2537         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
2538         * tree-complex.c (expand_complex_libcall): Pass cfun to
2539         stmt_could_throw_p and to stmt_can_throw_internal.
2540         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
2541         * tree-inline.c (copy_edges_for_bb): Likewise.
2542         (maybe_move_debug_stmts_to_successors): Likewise.
2543         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
2544         stmt_could_throw_p.
2545         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
2546         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
2547         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
2548         stmt_can_throw_internal.
2549         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
2550         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
2551         stmt_could_throw_p.
2552         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
2553         stmt_can_throw_internal.
2554         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
2555         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
2556         stmt_could_throw_p.
2557         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
2558         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
2559         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
2560         (convert_mult_to_fma_1): Likewise.
2561         (convert_to_divmod): Likewise.
2562         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
2563         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
2564         * tree-ssa-propagate.c
2565         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
2566         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
2567         (maybe_optimize_range_tests): Likewise.
2568         (linearize_expr_tree): Likewise.
2569         (reassociate_bb): Likewise.
2570         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2571         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
2572         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
2573         (handle_char_store): Likewise.
2574         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
2575         stmt_can_throw_internal.
2576         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
2577         stmt_could_throw_p.
2578         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2579         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
2580         (vectorizable_simd_clone_call): Likewise.
2581         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
2582         (gimple_stringop_fixed_value): Likewise.
2584 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
2586         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
2587         literal pool references.
2588         (s390_check_qrst_address): Adapt to the new behavior of
2589         s390_loadrelative_operand_p ().
2591 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
2593         PR target/72782
2594         * config/i386/sse.md (*andnot<mode>3_bcst): New.
2596 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
2598         PR target/72782
2599         * config/i386/sse.md (*<code><mode>3_bcst): New.
2601 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
2603         PR target/72782
2604         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
2605         V4DI, V16SI and V8DI.
2606         (*sub<mode>3<mask_name>_bcst): New.
2607         (*add<mode>3<mask_name>_bcst): Likewise.
2609 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
2610             Jinsong Ji  <jji@us.ibm.com>
2612         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
2613         __m64 with __vector unsigned long long for compatibility.
2614         (_mm_movemask_epi8): Likewise.
2615         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
2616         (_mm_cvttps_pi32): Likewise.
2617         (_mm_cvtpi32_ps): Likewise.
2618         (_mm_cvtps_pi16): Likewise.
2619         (_mm_loadh_pi): Likewise.
2620         (_mm_storeh_pi): Likewise.
2621         (_mm_movehl_ps): Likewise.
2622         (_mm_movelh_ps): Likewise.
2623         (_mm_loadl_pi): Likewise.
2624         (_mm_storel_pi): Likewise.
2625         (_mm_movemask_ps): Likewise.
2626         (_mm_shuffle_pi16): Likewise.
2628 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
2630         PR target/72782
2631         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
2632         __builtin_ia32_vfnmsubpd512_mask.
2633         (_mm512_mask_fnmsub_round_pd): Likewise.
2634         (_mm512_fnmsub_pd): Likewise.
2635         (_mm512_mask_fnmsub_pd): Likewise.
2636         (_mm512_maskz_fnmsub_round_pd): Use
2637         __builtin_ia32_vfnmsubpd512_maskz.
2638         (_mm512_maskz_fnmsub_pd): Likewise.
2639         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
2640         (_mm512_mask_fnmsub_round_ps): Likewise.
2641         (_mm512_fnmsub_ps): Likewise.
2642         (_mm512_mask_fnmsub_ps): Likewise.
2643         (_mm512_maskz_fnmsub_round_ps): Use
2644         __builtin_ia32_vfnmsubps512_maskz.
2645         (_mm512_maskz_fnmsub_ps): Likewise.
2646         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
2647         __builtin_ia32_vfnmsubpd256_mask.
2648         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
2649         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
2650         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
2651         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
2652         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
2653         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
2654         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
2655         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
2656         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
2657         __builtin_ia32_vfnmsubpd.
2658         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
2659         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
2660         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
2661         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
2662         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
2663         * config/i386/i386-builtin.def: Add
2664         __builtin_ia32_vfnmsubpd256_mask,
2665         __builtin_ia32_vfnmsubpd256_maskz,
2666         __builtin_ia32_vfnmsubpd128_mask,
2667         __builtin_ia32_vfnmsubpd128_maskz,
2668         __builtin_ia32_vfnmsubps256_mask,
2669         __builtin_ia32_vfnmsubps256_maskz,
2670         __builtin_ia32_vfnmsubps128_mask,
2671         __builtin_ia32_vfnmsubps128_maskz,
2672         __builtin_ia32_vfnmsubpd512_mask,
2673         __builtin_ia32_vfnmsubpd512_maskz,
2674         __builtin_ia32_vfnmsubps512_mask,
2675         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
2676         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
2677         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
2678         __builtin_ia32_vfnmsubpd256.
2679         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
2680         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
2681         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
2682         Likewise.
2683         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
2684         Likewise.
2685         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
2686         Likewise.
2687         (fmai_vmfnmsub_<mode><round_name>): Likewise.
2689 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
2691         PR target/72782
2692         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
2693         __builtin_ia32_vfnmaddpd512_mask.
2694         (_mm512_mask_fnmadd_round_pd): Likewise.
2695         (_mm512_fnmadd_pd): Likewise.
2696         (_mm512_mask_fnmadd_pd): Likewise.
2697         (_mm512_maskz_fnmadd_round_pd): Use
2698         __builtin_ia32_vfnmaddpd512_maskz.
2699         (_mm512_maskz_fnmadd_pd): Likewise.
2700         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
2701         (_mm512_mask_fnmadd_round_ps): Likewise.
2702         (_mm512_fnmadd_ps): Likewise.
2703         (_mm512_mask_fnmadd_ps): Likewise.
2704         (_mm512_maskz_fnmadd_round_ps): Use
2705         __builtin_ia32_vfnmaddps512_maskz.
2706         (_mm512_maskz_fnmadd_ps): Likewise.
2707         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
2708         __builtin_ia32_vfnmaddpd256_mask.
2709         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
2710         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
2711         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
2712         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
2713         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
2714         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
2715         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
2716         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
2717         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
2718         __builtin_ia32_vfnmaddpd.
2719         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
2720         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
2721         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
2722         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
2723         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
2724         * config/i386/i386-builtin.def: Add
2725         __builtin_ia32_vfnmaddpd256_mask,
2726         __builtin_ia32_vfnmaddpd256_maskz,
2727         __builtin_ia32_vfnmaddpd128_mask,
2728         __builtin_ia32_vfnmaddpd128_maskz,
2729         __builtin_ia32_vfnmaddps256_mask,
2730         __builtin_ia32_vfnmaddps256_maskz,
2731         __builtin_ia32_vfnmaddps128_mask,
2732         __builtin_ia32_vfnmaddps128_maskz,
2733         __builtin_ia32_vfnmaddpd512_mask,
2734         __builtin_ia32_vfnmaddpd512_maskz,
2735         __builtin_ia32_vfnmaddps512_mask,
2736         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
2737         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
2738         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
2739         __builtin_ia32_vfnmaddpd256.
2740         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
2741         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
2742         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
2743         Likewise.
2744         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
2745         Likewise.
2746         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
2747         Likewise.
2748         (fmai_vmfnmadd_<mode><round_name>): Likewise.
2750 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
2752         PR target/72782
2753         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
2754         __builtin_ia32_vfmsubpd512_mask.
2755         (_mm512_mask_fmsub_round_pd): Likewise.
2756         (_mm512_fmsub_pd): Likewise.
2757         (_mm512_mask_fmsub_pd): Likewise.
2758         (_mm512_maskz_fmsub_round_pd): Use
2759         __builtin_ia32_vfmsubpd512_maskz.
2760         (_mm512_maskz_fmsub_pd): Likewise.
2761         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
2762         (_mm512_mask_fmsub_round_ps): Likewise.
2763         (_mm512_fmsub_ps): Likewise.
2764         (_mm512_mask_fmsub_ps): Likewise.
2765         (_mm512_maskz_fmsub_round_ps): Use
2766         __builtin_ia32_vfmsubps512_maskz.
2767         (_mm512_maskz_fmsub_ps): Likewise.
2768         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
2769         __builtin_ia32_vfmsubpd256_mask.
2770         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
2771         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
2772         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
2773         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
2774         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
2775         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
2776         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
2777         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
2778         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
2779         __builtin_ia32_vfmsubpd.
2780         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
2781         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
2782         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
2783         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
2784         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
2785         * config/i386/i386-builtin.def: Add
2786         __builtin_ia32_vfmsubpd256_mask,
2787         __builtin_ia32_vfmsubpd256_maskz,
2788         __builtin_ia32_vfmsubpd128_mask,
2789         __builtin_ia32_vfmsubpd128_maskz,
2790         __builtin_ia32_vfmsubps256_mask,
2791         __builtin_ia32_vfmsubps256_maskz,
2792         __builtin_ia32_vfmsubps128_mask,
2793         __builtin_ia32_vfmsubps128_maskz,
2794         __builtin_ia32_vfmsubpd512_mask,
2795         __builtin_ia32_vfmsubpd512_maskz,
2796         __builtin_ia32_vfmsubps512_mask,
2797         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
2798         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
2799         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
2800         __builtin_ia32_vfmsubpd256.
2801         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
2802         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
2803         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
2804         Likewise.
2805         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
2806         Likewise.
2807         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
2808         Likewise.
2809         (fmai_vmfmsub_<mode><round_name>): Likewise.
2811 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
2813         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
2814         Remove plus.  Renamed to ...
2815         (*sub<mode>3<mask_name>_bcst): This.
2816         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
2817         (*add<mode>3<mask_name>_bcst): This.
2819 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
2821         PR target/72782
2822         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
2824 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
2826         PR target/87662
2827         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
2828         (_mm_or_epi32): Likewise.
2829         (_mm256_xor_epi32): Likewise.
2830         (_mm_xor_epi32): Likewise.
2831         (_mm256_or_epi64): Likewise.
2832         (_mm_or_epi64): Likewise.
2833         (_mm256_xor_epi64): Likewise.
2834         (_mm_xor_epi64): Likewise.
2836 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
2838         PR target/72782
2839         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
2841 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
2843         PR middle-end/87647
2844         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
2846 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
2848         * doc/ux.texi: Move @section directly after @node.
2850 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
2852         PR middle-end/85488
2853         PR middle-end/87649
2854         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
2855         depend closely nested inside of loop with ordered clause with
2856         a parameter.
2858 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
2860         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
2861         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
2862         * doc/ux.texi: New file.
2864 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
2866         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
2867         be the first CR field allocated.
2869 2018-10-19  Richard Biener  <rguenther@suse.de>
2871         PR target/87657
2872         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
2873         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
2875 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
2877         PR target/72782
2878         * config/i386/sse.md
2879         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
2880         (*add<mode>3<mask_name>_bcst_2): Likewise.
2882 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
2884         * config/i386/sse.md
2885         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
2886         Replace nonimmediate_operand with register_operand.
2887         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
2888         Likewise.
2889         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
2890         Likewise.
2892 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
2894         PR rtl-optimization/87596
2895         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
2896         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
2897         for instructions in FROM..TO range.
2899 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
2901         * cfgexpand.c (expand_one_var): Use specific wording in error message
2902         for non-local frame variables.
2903         * stor-layout.c (layout_decl): Do not issue a warning for them.
2905 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
2907         * haifa-sched.c (priority): Add force_recompute parameter.
2908         (apply_replacement): Call priority () with force_recompute = true.
2909         (restore_pattern): Likewise.
2911 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
2913         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
2914         HOST_BITS_PER_WIDE_INT.
2915         (test_vector_ops_duplicate): Likewise.
2917 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
2919         PR target/72782
2920         * config/i386/sse.md (VF_AVX512): New.
2921         (avx512bcst): Likewise.
2922         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
2923         Likewise.
2924         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
2925         Likewise.
2926         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
2927         Likewise.
2929 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
2931         * doc/invoke.texi (-dumpversion): Improve grammar.
2932         (-dumpfullversion): Make more consistent with -dumpversion.
2934 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
2936         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
2937         Set JUMP_LABEL to the jump insn.
2938         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
2939         Predict emitted jump and add label to jump insn.
2941 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
2943         PR tree-optimization/87562
2944         * input.c (get_substring_ranges_for_loc): Use
2945         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
2946         getting the linemap for the endpoint.  Verify that it's either
2947         in the same linemap as the start point's spelling location, or
2948         at least in the same file.
2950 2018-10-18  Richard Biener  <rguenther@suse.de>
2952         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
2953         feed width-specific load/store costs through ix86_vec_cost.
2954         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
2955         (k8_cost): Likewise.
2956         (bdver_cost): Likewise.
2957         (znver1_cost): Likewise.
2958         (btver1_cost): Likewise.
2959         (btver2_cost): Likewise.
2961 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
2963         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
2964         to simplify subreg of vec_merge.
2966 2018-10-18  Richard Biener  <rguenther@suse.de>
2968         * config/i386/i386.c: Fix costing of vector FMA.
2970 2018-10-18  Richard Biener  <rguenther@suse.de>
2972         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
2973         and argument.
2974         (ix86_builtin_vectorization_cost): For vec_construct properly
2975         cost insertion into SSE regs.
2976         (...): Adjust calls to ix86_vec_cost.
2978 2018-10-18  Richard Biener  <rguenther@suse.de>
2980         PR middle-end/87087
2981         Revert
2982         2018-02-07  Richard Biener  <rguenther@suse.de>
2984         PR tree-optimization/84204
2985         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2986         this place.
2988 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
2990         PR target/87537
2991         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
2992         of vec_duplicate.
2993         (test_vector_ops_duplicate): Add test for a scalar subreg of a
2994         VEC_MERGE of a VEC_DUPLICATE.
2996 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
2998         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
2999         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
3000         * doc/standards.texi (C Language): Document C2X.
3001         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
3002         (rl78_option_override): Handle "GNU C2X" language name.
3004 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
3006         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
3007         Document C17 as published in 2018.
3009 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
3011         PR middle-end/87623
3012         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
3013         bail out if both sides do not have the same storage order.
3015 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
3017         * bitmap.c (bitmap_head::dump): New.
3018         * bitmap.h (bitmap_head): Add dump().
3019         * gimple-ssa-evrp-analyze.c
3020         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
3021         (evrp_range_analyzer::set_ssa_range_info): Same.
3022         (evrp_range_analyzer::record_ranges_from_phis): Same.
3023         (evrp_range_analyzer::record_ranges_from_stmt): Same.
3024         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
3025         * gimple-ssa-sprintf.c (get_int_range): Same.
3026         (format_integer): Same.
3027         (sprintf_dom_walker::handle_gimple_call): Same.
3028         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
3029         (ipcp_vr_lattice::top_p): Same.
3030         (ipcp_vr_lattice::bottom_p): Same.
3031         (ipcp_vr_lattice::set_to_bottom): Same.
3032         (ipa_vr_operation_and_type_effects): Same.
3033         (propagate_vr_across_jump_function): Same.
3034         (ipcp_store_vr_results): Same.
3035         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
3036         (ipa_print_node_jump_functions_for_edge): Same.
3037         (ipa_get_value_range): Same.
3038         (ipa_compute_jump_functions_for_edge): Same.
3039         (ipa_write_jump_function): Same.
3040         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
3041         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
3042         Same.
3043         * vr-values.c (set_value_range_to_nonnegative): Same.
3044         (set_value_range_to_truthvalue): Same.
3045         (vr_values::get_value_range): Same.
3046         (vr_values::set_defs_to_varying): Same.
3047         (vr_values::update_value_range): Same.
3048         (symbolic_range_based_on_p): Same.
3049         (vr_values::op_with_boolean_value_range_p): Same.
3050         (vr_values::extract_range_for_var_from_comparison_expr): Same.
3051         (vr_values::extract_range_from_ssa_name): Same.
3052         (vr_values::extract_range_from_binary_expr): Same.
3053         (vr_values::extract_range_from_unary_expr): Same.
3054         (vr_values::extract_range_from_cond_expr): Same.
3055         (vr_values::extract_range_from_comparison): Same.
3056         (vr_values::check_for_binary_op_overflow): Same.
3057         (vr_values::extract_range_basic): Same.
3058         (vr_values::extract_range_from_assignment): Same.
3059         (compare_ranges): Same.
3060         (compare_range_with_value): Same.
3061         (vr_values::adjust_range_with_scev): Same.
3062         (vrp_valueize): Same.
3063         (vrp_valueize_1): Same.
3064         (vr_values::get_vr_for_comparison): Same.
3065         (vr_values::compare_name_with_value): Same.
3066         (vr_values::compare_names): Same.
3067         (vr_values::vrp_evaluate_conditional): Same.
3068         (find_case_label_ranges): Same.
3069         (vr_values::vrp_visit_switch_stmt): Same.
3070         (vr_values::extract_range_from_phi_node): Same.
3071         (vr_values::simplify_div_or_mod_using_ranges): Same.
3072         (vr_values::simplify_bit_ops_using_ranges): Same.
3073         (test_for_singularity): Same.
3074         (range_fits_type_p): Same.
3075         (vr_values::simplify_cond_using_ranges_1): Same.
3076         (vr_values::simplify_switch_using_ranges): Same.
3077         (vr_values::simplify_float_conversion_using_ranges): Same.
3078         (vr_values::two_valued_val_range_p): Same.
3079         (vr_values::add_equivalence): Move to value_range::equiv_add.
3080         * vr-values.h (vr_values::add_equivalence): Remove.
3081         (VR_INITIALIZER): Remove.
3082         * tree-vrp.c (value_range::set): New.
3083         (value_range::equiv_add): New.
3084         (value_range::value_range): New.
3085         (value_range::deep_copy): New.
3086         (value_range::check): New.
3087         (value_range::equal_p): New.
3088         (value_range::ignore_equivs_equal_p): New.
3089         (value_range::operator==): New.
3090         (value_range::operator!=): New.
3091         (value_range::symbolic_p): New.
3092         (value_range::numeric_p): New.
3093         (value_range::set_undefined): New.
3094         (value_range::set_varying): New.
3095         (value_range::may_contain_p): New.
3096         (value_range::equiv_clear): New.
3097         (value_range::singleton_p): New.
3098         (value_range::intersect): New.
3099         (value_range::dump): New.
3100         (value_range::set_and_canonicalize): New.
3101         (set_value_range): Adjust for value_range API.
3102         (set_value_range_to_undefined): Same.
3103         (set_value_range_to_varying): Same.
3104         (set_and_canonicalize_value_range): Same.
3105         (set_value_range_to_nonnull): Same.
3106         (set_value_range_to_null): Same.
3107         (range_is_null): Same.
3108         (range_is_nonnull): Same.
3109         (range_int_cst_p): Same.
3110         (range_int_cst_singleton_p): Same.
3111         (symbolic_range_p): Same.
3112         (range_includes_zero_p): Same.
3113         (value_range_constant_singleton): Same.
3114         (vrp_set_zero_nonzero_bits): Same.
3115         (ranges_from_anti_range): Same.
3116         (extract_range_into_wide_ints): Same.
3117         (extract_range_from_multiplicative_op): Same.
3118         (set_value_range_with_overflow): Same.
3119         (extract_range_from_binary_expr_1): Same.
3120         (extract_range_from_unary_expr): Same.
3121         (dump_value_range): Same.
3122         (debug_value_range): Same.
3123         (vrp_prop::check_array_ref): Same.
3124         (vrp_prop::check_mem_ref): Same.
3125         (vrp_prop::vrp_initialize): Same.
3126         (vrp_prop::visit_stmt): Same.
3127         (intersect_ranges): Same.
3128         (vrp_prop::visit_phi): Same.
3129         (vrp_prop::vrp_finalize): Same.
3130         (determine_value_range_1): Same.
3131         (determine_value_range): Same.
3132         (vrp_intersect_ranges_1): Rename to...
3133         (vrp_intersect_1): this.
3134         (vrp_intersect_ranges): Rename to...
3135         (value_range::intersect_helper): ...this.
3136         (vrp_meet_1): Rename to...
3137         (value_range::union_helper): ...this.
3138         (vrp_meet): Rename to...
3139         (value_range::union_): ...this.
3140         (copy_value_range): Remove.
3141         * tree-vrp.h (struct value_range): Rewrite into a proper class.
3142         (value_range::vrtype): New.
3143         (value_range::type): New.
3144         (value_range::equiv): New.
3145         (value_range::min): New.
3146         (value_range::max): New.
3147         (value_range::varying_p): New.
3148         (value_range::undefined_p): New.
3149         (value_range::null_p): New.
3150         (value_range::equiv_add): New.
3151         (copy_value_range): Remove.
3153 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
3155         * Makefile.in (SELFTEST_TARGETS): New.
3156         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
3157         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
3158         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
3159         c/Make-lang.in.
3160         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
3161         (selftest-c++-gdb, selftest-c++-valgrind): Move to
3162         cp/Make-lang.in.
3163         * configure: Regenerate.
3164         * configure.ac (selftest_languages): New.
3166 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
3168         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
3169         overflow wraps argument.
3170         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
3171         wide_int_range_multiplicative_op.
3172         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
3173         overflow wraps argument.
3174         (wide_int_range_multiplicative_op): Same.
3175         (wide_int_range_lshift): Same.
3176         (wide_int_range_div): Same.
3177         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
3178         (wide_int_range_lshift): Same.
3179         (wide_int_range_div): Same.
3181 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
3183         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
3184         use sign as argument.
3185         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
3186         wide_int_range_shift_undefined_p.
3188 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3190         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
3191         Rename to...
3192         (@despeculate_copy<ALLI_TI:mode>): ... This.
3193         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
3194         switch statement.
3196 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3198         * config.gcc: Obsolete *-*-solaris2.10*.
3199         * doc/install.texi (Specific, *-*-solaris2*): Document it.
3201 2018-10-12  Jeff Law  <law@redhat.com>
3203         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
3204         reg + sym +- const_int addressing modes.
3206 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
3208         * common.opt (fdiagnostics-minimum-margin-width=): New option.
3209         * diagnostic-show-locus.c (layout::layout): Apply the minimum
3210         margin width.
3211         (layout::start_annotation_line): Only print up to 3 of the
3212         margin character, to avoid touching the left-hand side.
3213         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
3214         minimum margin width, as set by test_diagnostic_context's ctor.
3215         (selftest::test_fixit_insert_containing_newline): Likewise.
3216         (selftest::test_fixit_insert_containing_newline_2): Likewise.
3217         (selftest::test_line_numbers_multiline_range): Clear
3218         dc.min_margin_width.
3219         * diagnostic.c (diagnostic_initialize): Initialize
3220         min_margin_width.
3221         * diagnostic.h (struct diagnostic_context): Add field
3222         "min_margin_width".
3223         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
3224         * opts.c (common_handle_option): Handle
3225         OPT_fdiagnostics_minimum_margin_width_.
3226         * selftest-diagnostic.c
3227         (selftest::test_diagnostic_context::test_diagnostic_context):
3228         Initialize min_margin_width to 6.
3229         * toplev.c (general_init): Initialize global_dc->min_margin_width.
3231 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
3233         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3234         Fix usage of "error_at_rich_loc" in the comment.
3236 2018-10-15  Renlin Li  <renlin.li@arm.com>
3238         PR target/87563
3239         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
3240         if-conversioned loop when it contains ifn with types not
3241         supported by backend.
3242         * internal-fn.c (expand_direct_optab_fn): Add an assert.
3243         (direct_internal_fn_supported_p): New helper function.
3244         * internal-fn.h (direct_internal_fn_supported_p): Declare.
3246 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
3248         PR target/87572
3249         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
3250         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
3251         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
3253 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
3255         PR tree-optimization/87022
3256         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
3257         bits in dist vector rather than the first one.
3259 2018-10-15  Richard Biener  <rguenther@suse.de>
3261         PR middle-end/87610
3262         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
3263         (visit_loadstore): When a used restrict tag escaped verify that
3264         the points-to solution of "other" pointers do not include
3265         escaped.
3266         (compute_dependence_clique): If a used restrict tag escaped
3267         communicated that down to visit_loadstore.
3269 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
3271         * config/s390/s390.c (s390_expand_vec_init): Force vector element
3272         into reg if it isn't a general operand.
3274 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
3276         PR target/87599
3277         * config/i386/sse.md (*vec_dupv2di): Add register source to
3278         movddup.
3280 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
3282         PR target/87572
3283         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
3284         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
3285         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
3286         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
3288 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
3290         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
3291         (notice_args_size): Set it in the current trace if no insn that can
3292         throw internally has been seen yet.
3293         (connect_traces): When connecting args_size between traces, allow the
3294         incoming values not to match if there is an insn setting it before the
3295         first insn that can throw internally; in that case, force the creation
3296         of a CFI note on this latter insn.
3298 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
3300         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
3301         the base class, because using the injected-class-name was not clearly
3302         specified until DR 176.
3304 2018-10-12  Paul Koning  <ni1d@arrl.net>
3306         * config/pdp11/pdp11.md (doloop_end): New expander.
3307         (doloop_end_insn): renamed from "doloop_end".
3308         (addqi3): New pattern.
3309         (subqi3): New pattern.
3310         * config/pdp11/predicates.md (incdec_operand): New predicate.
3312 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
3314         PR middle-end/81376
3315         * real.c (format_helper::can_represent_integral_type_p): New function
3316         * real.h (format_helper::can_represent_integral_type_p): Ditto.
3317         * match.pd: New pattern.
3319 2018-10-12  Alexandre Oliva <oliva@adacore.com>
3321         * configure.ac: Introduce --enable-large-address-aware
3322         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
3323         * doc/install.texi: Document it.
3324         * configure, config.in: Rebuilt.
3325         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
3326         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
3327         (LINK_SPEC): Insert it.
3328         * config/i386/mingw-w64.h: Likewise.
3330         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
3332 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
3334         PR rtl-optimization/87600
3335         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
3337 2018-10-12  Paul Koning  <ni1d@arrl.net>
3339         * doc/md.texi (doloop_end): Document that the pattern code may
3340         need to check operand mode.
3342 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
3344         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
3345         to zero-extend between int and floating-point registers.
3346         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
3347         ldp into floating-point registers.  Add type and arch attributes.
3348         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
3349         Use f_loads for type attribute.
3351 2018-10-11  Martin Sebor  <msebor@redhat.com>
3353         * doc/extend.texi (attribute packed): Correct typos.
3355 2018-10-11  Martin Sebor  <msebor@redhat.com>
3357         * doc/extend.texi (attribute flatten): Mention interaction with
3358         noinline.
3360 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
3362         PR target/87156
3363         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3364         Set new_decl virtual flag to zero.
3366 2018-10-11  Martin Sebor  <msebor@redhat.com>
3368         PR middle-end/87593
3369         * doc/extend.texi (attribute format_arg): Discuss using multiple
3370         attributes on a single function.
3372 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
3374         PR tree-optimization/86829
3375         * match.pd (sin (atan (x))): New simplification rules.
3376         (cos (atan (x))): Likewise.
3377         * real.c (build_sinatan_real): New function.
3378         * real.h (build_sinatan_real): Prototype.
3380 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
3382         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
3383         function.
3384         (fold_mergeeo_helper): New helper function.
3385         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
3386         intrinsics.  Correct some whitespace indentation issues.
3388 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
3390         PR target/87511
3391         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
3392         Use HOST_WIDE_INT_1U for shift.
3394 2018-10-11  Doug Rupp  <rupp@adacore.com>
3395             Olivier Hainque  <hainque@adacore.com>
3397         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
3398         Pass --relax to the linker for RTPs.
3399         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
3401 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
3402             Jan Hubicka  <jh@suse.cz>
3403             Martin Jambor  <mjambor@suse.cz>
3405         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
3406         the same elements are repeated rather than printing all of them.
3407         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
3408         "repeated" elements.
3409         * read-rtl-function.c (test_loading_repeat): New function.
3410         (read_rtl_function_c_tests): Call test_loading_repeat.
3411         * rtl-tests.c (test_dumping_repeat): New function.
3412         (rtl_tests_c_tests): Call test_dumping_repeat.
3414 2018-10-11  Richard Biener  <rguenther@suse.de>
3416         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
3417         bdver?_cost): Unify to ...
3418         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
3419         * config/i386/i386.c (processor_cost_table): Adjust.
3421 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
3423         PR middle-end/87574
3424         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
3425         the thunk when expanding to GIMPLE.
3427 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3429         * varasm.c (mergeable_string_section): Don't try to move zero-length
3430         strings to the merge section.
3432 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
3434         PR target/87573
3435         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
3437 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
3439         PR target/87550
3440         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
3441         to special_args set.
3443 2018-10-10  Richard Biener  <rguenther@suse.de>
3445         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
3446         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
3447         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
3448         and recursing and pattern terminating the recursion on SSE
3449         vector width using ix86_expand_reduc.
3450         (reduc_sminmax_scal_<mode>): Split into part reducing to half
3451         width and recursing and SSE2 vector variant doing the final
3452         reduction with ix86_expand_reduc.
3453         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
3454         with terminating the recursion at AVX level, splitting that
3455         to SSE there.
3457 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
3459         * genmatch.c (error_cb): Rename to...
3460         (diagnostic_cb): ...this, converting int params to enums.
3461         (fatal_at): Update for renaming.
3462         (warning_at): Likewise.
3463         (main): Likewise.
3464         * input.c (selftest::ebcdic_execution_charset::apply):
3465         Update for renaming of...
3466         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
3467         to...
3468         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
3469         converting level and reason to enums.
3470         (class selftest::lexer_error_sink): Rename to...
3471         (class selftest::lexer_test_options): ...this, renaming field
3472         "m_errors" to "m_diagnostics".
3473         (selftest::lexer_test_options::apply): Update for renaming of...
3474         (selftest::lexer_test_options::on_error): ...this, renaming to...
3475         (selftest::lexer_test_options::on_diagnostic): ...this
3476         converting level and reason to enums.
3477         (selftest::test_lexer_string_locations_raw_string_unterminated):
3478         Update for renamings.
3479         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
3480         "reason".
3482 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
3484         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
3485         * config/rs6000/pmmintrin.h: New file.
3487 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
3489         PR tree-optimization/86659
3490         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
3492 2018-10-09  Richard Biener  <rguenther@suse.de>
3494         PR tree-optimization/63155
3495         * tree-ssa-structalias.c: Include tree-ssa.h.
3496         (get_constraint_for_ssa_var): For undefs return nothing_id.
3497         (find_func_aliases): Cleanup PHI handling.
3499 2018-10-09  Richard Biener  <rguenther@suse.de>
3501         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
3502         replacements.
3504 2018-10-09  Martin Liska  <mliska@suse.cz>
3506         * asan.c (asan_emit_stack_protection): If a stack variable
3507         is located in a same file as current function, then emit
3508         line info into variable definition string.
3510 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
3512         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
3513         information.
3515 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
3517         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
3518         on the thunk.
3520 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
3522         PR tree-optimization/86659
3523         * gimple-match.h (struct gimple_match_op): Add reverse field.
3524         (gimple_match_op::set_op): New overloaded method.
3525         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
3526         the REF_REVERSE_STORAGE_ORDER flag on the value.
3527         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
3528         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
3529         
3530 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
3532         PR middle-end/63155
3533         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
3534         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
3536 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
3538         PR target/87517
3539         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
3540         Defined with __builtin_ia32_vfmaddsubpd512_mask.
3542 2018-10-08  Richard Biener  <rguenther@suse.de>
3544         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
3545         cost the same as AVX128 ones.
3547 2018-10-08  Paul Koning  <ni1d@arrl.net>
3549         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
3550         (expand_block_move): New function.
3551         * config/pdp11/pdp11.c (output_block_move): Remove.
3552         (expand_block_move): New function.
3553         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
3554         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
3555         (*movmemhi1): Remove.
3557 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
3559         * config/s390/2827.md: Increase latencies for some FP instructions.
3561 2018-10-08  Richard Biener  <rguenther@suse.de>
3563         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3564         Open a dump scope.
3565         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
3566         * tree-vectorizer.h (dump_stmt_cost): Adjust.
3567         (add_stmt_cost): Dump return value of the hook.
3569 2018-10-08  Richard Biener  <rguenther@suse.de>
3571         PR tree-optimization/63155
3572         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
3573         (ssa_propagation_engine::ssa_propagate): Remove redundant
3574         bitmap bit clearing.
3576 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
3578         PR rtl-optimization/86939
3579         PR rtl-optimization/87479
3580         * ira.h (non_conflicting_reg_copy_p): New prototype.
3581         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
3582         (make_hard_regno_dead): Don't add conflicts for register
3583         ignore_reg_for_conflicts.
3584         (make_object_dead): Likewise.
3585         (non_conflicting_reg_copy_p): New function.
3586         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
3587         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
3588         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
3589         (make_hard_regno_dead): Don't add conflicts for register
3590         ignore_reg_for_conflicts.  Remove special conflict handling of
3591         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
3592         check_pic_pseudo_p and update callers.
3593         (mark_pseudo_dead): Don't add conflicts for register
3594         ignore_reg_for_conflicts.
3595         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
3597 2018-10-05  Andrew Waterman  <andrew@sifive.com>
3598             Jim Wilson  <jimw@sifive.com>
3600         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
3601         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
3602         new pattern using HONOR_SNANS that emits one extra instruction.
3604 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
3606         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
3607         patterns): Merge SI and DI patterns to a GPR pattern.
3608         (unnamed define_insn and define_split for record form of that): Merge
3609         to a single define_insn_and_split pattern.
3611 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
3613         PR c++/56856
3614         * input.c (expand_location_to_spelling_point): Add param "aspect"
3615         and use rather than hardcoding LOCATION_ASPECT_CARET.
3616         (get_substring_ranges_for_loc): Handle the case of a single token
3617         within a macro expansion.
3618         * input.h (expand_location_to_spelling_point): Add "aspect" param,
3619         defaulting to LOCATION_ASPECT_CARET.
3621 2018-10-05  Paul Koning  <ni1d@arrl.net>
3623         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
3624         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
3625         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
3626         (pdp11_guard_type): New function.
3628 2018-10-05  Paul Koning  <ni1d@arrl.net>
3630         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
3631         * config/pdp11/pdp11.opt (mfloat32): Remove.
3632         (mfloat64): Remove.
3633         * doc/invoke.texi (pdp11 -mfloat32): Remove:
3634         (pdp11 -mfloat64): Remove.
3636 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
3638         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
3639         (*cmp<mode>_cc_i387): Ditto.
3640         (*cmpu<mode>_cc_i387): Ditto.
3641         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
3642         * config/i386/i386.c (ix86_expand_fp_compare): Remove
3643         "scratch" argument.
3644         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
3645         Emit x86_sahf_1 pattern.
3646         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
3647         (ix86_expand_carry_flag_compare): Ditto.
3649 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
3651         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
3652         to reg_or_0_operand.  Add "C" constraint.
3653         (*cmpxf_cc_i387): Ditto.
3654         (*cmp<mode>_i387): Change operand 2 predicate
3655         to nonimm_or_0_operand.  Add "C" constraint.
3656         (*cmp<mode>_cc_i387): Ditto.
3657         (*cmp<mode>_0_i387): Remove insn pattern.
3658         (*cmp<mode>_0_cc_i387): Ditto.
3660 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
3662         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
3663         * config/i386/predicates.md (nonimm_or_0_operand): Rename
3664         from vector_move_operand.  Update all uses.
3666 2018-10-05  Martin Sebor  <msebor@redhat.com>
3668         PR tree-optimization/87490
3669         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
3670         consistently.
3672 2018-10-05  Richard Biener  <rguenther@suse.de>
3674         PR tree-optimization/63155
3675         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
3676         vertical space in dumpfiles.
3677         * tree-ssa-propagate.h
3678         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
3679         * tree-ssa-propagate.c (cfg_blocks_back): New global.
3680         (ssa_edge_worklist_back): Likewise.
3681         (curr_order): Likewise.
3682         (cfg_blocks_get): Remove abstraction.
3683         (cfg_blocks_add): Likewise.
3684         (cfg_blocks_empty_p): Likewise.
3685         (add_ssa_edge): Add to current or next worklist based on
3686         RPO index.
3687         (add_control_edge): Likewise.
3688         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
3689         into ...
3690         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
3691         iteration from CFG and SSA edge worklist so we process
3692         everything in RPO order, prioritizing forward progress
3693         over iteration.
3694         (ssa_prop_init): Allocate new worklists, do not dump
3695         immediate uses.
3696         (ssa_prop_fini): Free new worklists.
3698 2018-10-05  Richard Biener  <rguenther@suse.de>
3700         * tree-core.h (tree_block::abstract_flag): Remove.
3701         (tree_block::block_num): Make full 32bits.
3702         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
3703         * tree.h (BLOCK_ABSTRACT): Remove.
3704         * dwarf2out.c (gen_lexical_block_die): Remove dead code
3705         resulting from BLOCK_ABSTRACT being always false.
3706         (gen_inlined_subroutine_die): Likewise.
3707         (gen_block_die): Likewise.
3708         * tree.c (block_ultimate_origin): Likewise.
3709         * tree-pretty-print.c (dump_block_node): Remove code dealing
3710         with BLOCK_ABSTRACT.
3711         * tree-ssa-live.c (dump_scope_block): Likewise.
3712         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
3713         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
3715 2018-10-05   Richard Biener  <rguenther@suse.de>
3717         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
3718         is asked for initialize mode to the component mode of the
3719         vector type.
3721 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
3723         PR target/87522
3724         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
3725         assembler for -mavx.
3726         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
3728 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
3730         PR target/87509
3731         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
3732         RS6000_BTM_DFP.
3733         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
3734         to be DImode.  When using mffscrn, force the operand to a register.
3736 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
3738         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
3739         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
3740         X87MODEF mode iterator.
3741         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
3742         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
3743         X87MODEF mode iterator.
3745 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
3747         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
3748         -Wno-prio-ctor-dtor.
3750 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
3752         * Makefile.in (OBJS): Add opt-problem.o.
3753         * dump-context.h: Include "selftest.h.
3754         (selftest::temp_dump_context): New forward decl.
3755         (class dump_context): Make friend of class
3756         selftest::temp_dump_context.
3757         (dump_context::dump_loc_immediate): New decl.
3758         (class dump_pretty_printer): Move here from dumpfile.c.
3759         (class temp_dump_context): Move to namespace selftest.
3760         (temp_dump_context::temp_dump_context): Add param
3761         "forcibly_enable_dumping".
3762         (selftest::verify_dumped_text):
3763         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
3764         (selftest::verify_item):
3765         (ASSERT_IS_TEXT): Move here from dumpfile.c.
3766         (ASSERT_IS_TREE): Likewise.
3767         (ASSERT_IS_GIMPLE): Likewise.
3768         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
3769         to...
3770         (dump_context::dump_loc_immediate): ...this new function.
3771         (class dump_pretty_printer): Move to dump-context.h.
3772         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
3773         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
3774         (temp_dump_context::temp_dump_context): Move to "selftest"
3775         namespace.  Add param "forcibly_enable_dumping", and use it to
3776         conditionalize the use of m_pp;
3777         (selftest::verify_dumped_text): Make non-static.
3778         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
3779         (selftest::verify_item): Make non-static.
3780         (ASSERT_IS_TEXT): Move to dump-context.h.
3781         (ASSERT_IS_TREE): Likewise.
3782         (ASSERT_IS_GIMPLE): Likewise.
3783         (selftest::test_capture_of_dump_calls): Pass "true" for new
3784         param of temp_dump_context.
3785         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
3786         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
3787         TDF_COMPARE_DEBUG.
3788         * opt-problem.cc: New file.
3789         * opt-problem.h: New file.
3790         * optinfo-emit-json.cc
3791         (selftest::test_building_json_from_dump_calls): Pass "true" for
3792         new param of temp_dump_context.
3793         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
3794         (optinfo::emit_for_opt_problem): New function.
3795         (optinfo::emit): Clarity which emit_item is used.
3796         * optinfo.h (optinfo::get_dump_location): New accessor.
3797         (optinfo::emit_for_opt_problem): New decl.
3798         (optinfo::emit): Make const.
3799         * selftest-run-tests.c (selftest::run_tests): Call
3800         selftest::opt_problem_cc_tests.
3801         * selftest.h (selftest::opt_problem_cc_tests): New decl.
3802         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
3803         bool to opt_result, converting fprintf messages to
3804         opt_result::failure_at calls.  Add "stmt" param for use by the
3805         failure_at calls.
3806         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
3807         (runtime_alias_check_p): Convert return type from bool to
3808         opt_result, converting dump_printf calls to
3809         opt_result::failure_at, using the statement DDR_A for their
3810         location.
3811         (find_data_references_in_stmt): Convert return type from bool to
3812         opt_result, converting "return false" to opt_result::failure_at
3813         with a new message.
3814         * tree-data-ref.h: Include "opt-problem.h".
3815         (dr_analyze_innermost): Convert return type from bool to opt_result,
3816         and add a const gimple * param.
3817         (find_data_references_in_stmt): Convert return type from bool to
3818         opt_result.
3819         (runtime_alias_check_p): Likewise.
3820         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
3821         dr_analyze_innermost.
3822         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
3823         Convert return type from bool to opt_result, adding a message for
3824         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
3825         (vect_analyze_data_ref_dependence): Convert return type from bool
3826         to opt_result.  Change sense of return type from "false"
3827         effectively meaning "no problems" to "false" meaning a problem,
3828         so that "return false" becomes "return opt_result::success".
3829         Convert "return true" calls to opt_result::failure_at, using
3830         the location of statement A rather than vect_location.
3831         (vect_analyze_data_ref_dependences): Convert return type from bool
3832         to opt_result.
3833         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
3834         calls to opt_result::failure_at, using the stmt location rather
3835         than vect_location.
3836         (vect_verify_datarefs_alignment): Convert return type from bool
3837         to opt_result.
3838         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
3839         into multiple more-tightly-scoped copies.
3840         (vect_analyze_data_refs_alignment): Convert return type from bool
3841         to opt_result.
3842         (vect_analyze_data_ref_accesses): Likewise, converting a
3843         "return false" to a "return opt_result::failure_at", adding a
3844         new message.
3845         (vect_prune_runtime_alias_test_list): Convert return type from
3846         bool to opt_result, converting dump_printf_loc to
3847         opt_result::failure_at.  Add a %G to show the pertinent statement,
3848         and use the stmt's location rather than vect_location.
3849         (vect_find_stmt_data_reference): Convert return type from
3850         bool to opt_result, converting dump_printf_loc to
3851         opt_result::failure_at, using stmt's location.
3852         (vect_analyze_data_refs):  Convert return type from bool to
3853         opt_result.  Convert "return false" to "return
3854         opt_result::failure_at", adding messages as needed.
3855         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
3856         type from bool to opt_result.
3857         (vect_determine_vf_for_stmt): Likewise.
3858         (vect_determine_vectorization_factor): Likewise, converting
3859         dump_printf_loc to opt_result::failure_at, using location of phi
3860         rather than vect_location.
3861         (vect_analyze_loop_form_1): Convert return type from bool to
3862         opt_result, converting dump_printf_loc calls, retaining the use of
3863         vect_location.
3864         (vect_analyze_loop_form): Convert return type from loop_vec_info
3865         to opt_loop_vec_info.
3866         (vect_analyze_loop_operations): Convert return type from bool to
3867         opt_result, converting dump_printf_loc calls, using the location
3868         of phi/stmt rather than vect_location where available.  Convert
3869         various "return false" to "return opt_result::failure_at" with
3870         "unsupported phi" messages.
3871         (vect_get_datarefs_in_loop): Convert return type from bool to
3872         opt_result.  Add a message for the
3873         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
3874         (vect_analyze_loop_2): Convert return type from bool to
3875         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
3876         each "goto again;", adding new messages where needed.
3877         Add "unsupported grouped {store|load}" messages.
3878         (vect_analyze_loop): Convert return type from loop_vec_info to
3879         opt_loop_vec_info.
3880         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
3881         bool to opt_result.
3882         * tree-vect-stmts.c (process_use): Likewise, converting
3883         dump_printf_loc call and using stmt location, rather than
3884         vect_location.
3885         (vect_mark_stmts_to_be_vectorized): Likeise.
3886         (vect_analyze_stmt): Likewise, adding a %G.
3887         (vect_get_vector_types_for_stmt): Convert return type from bool to
3888         opt_result, converting dump_printf_loc calls and using stmt
3889         location, rather than vect_location.
3890         (vect_get_mask_type_for_stmt): Convert return type from tree to
3891         opt_tree, converting dump_printf_loc calls and using stmt location.
3892         * tree-vectorizer.c: Include "opt-problem.h.
3893         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
3894         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
3895         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
3896         enabled, use it to report at the top level "couldn't vectorize
3897         loop" followed by the problem.
3898         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
3899         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
3900         to opt_result.
3901         (vect_analyze_stmt): Likewise.
3902         (vect_get_vector_types_for_stmt): Likewise.
3903         (tree vect_get_mask_type_for_stmt): Likewise.
3904         (vect_analyze_data_ref_dependences): Likewise.
3905         (vect_enhance_data_refs_alignment): Likewise.
3906         (vect_analyze_data_refs_alignment): Likewise.
3907         (vect_verify_datarefs_alignment): Likewise.
3908         (vect_analyze_data_ref_accesses): Likewise.
3909         (vect_prune_runtime_alias_test_list): Likewise.
3910         (vect_find_stmt_data_reference): Likewise.
3911         (vect_analyze_data_refs): Likewise.
3912         (vect_analyze_loop): Convert return type from loop_vec_info to
3913         opt_loop_vec_info.
3914         (vect_analyze_loop_form): Likewise.
3915         (vect_analyze_slp): Convert return type from bool to opt_result.
3917 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
3919         * doc/invoke.texi (-fopt-info): Document new "internals"
3920         sub-option.
3921         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
3922         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
3923         MSG_ALL_KINDS.
3924         (optinfo_verbosity_options): Add "internals".
3925         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
3926         (dump_context::apply_dump_filter_p): New member function.
3927         (dump_context::dump_loc): Use apply_dump_filter_p rather than
3928         explicitly masking the dump_kind.
3929         (dump_context::begin_scope): Increment the scope depth first.  Use
3930         apply_dump_filter_p rather than explicitly masking the dump_kind.
3931         (dump_context::emit_item): Use apply_dump_filter_p rather than
3932         explicitly masking the dump_kind.
3933         (dump_dec): Likewise.
3934         (dump_hex): Likewise.
3935         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
3936         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
3937         (opt_info_switch_p): Update handling of default
3938         MSG_OPTIMIZED_LOCATIONS to cope with default of
3939         MSG_PRIORITY_USER_FACING.
3940         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
3941         masking the dump_kind.
3942         (selftest::test_capture_of_dump_calls): Update test_dump_context
3943         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
3944         than MSG_ALL.  Generalize scope test to be run at all four
3945         combinations of with/without MSG_PRIORITY_USER_FACING and
3946         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
3947         for each of the two values.
3948         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
3949         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
3950         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
3951         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
3952         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
3953         with MSG_PRIORITY_*.
3954         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
3955         dump messages as MSG_PRIORITY_USER_FACING.
3956         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
3957         about the interaction with MSG_PRIORITY_*.
3959 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3961         * varasm.c (output_constant): Add new parameter merge_strings.
3962         Make strings properly zero terminated in merge string sections.
3963         (mergeable_string_section): Don't fail if the last char is non-zero.
3964         (assemble_variable_contents): Handle merge string sections.
3965         (assemble_variable): Likewise.
3966         (assemble_constant_contents): Likewise.
3967         (output_constant_def_contents): Likewise.
3968         (output_constructor_array_range,
3969         output_constructor_regular_field): Adjust call to output_constant.
3970         (output_object_block): Adjust call to assemble_constant_contents
3971         and assemble_variable_contents.
3973 2018-10-04  Martin Liska  <mliska@suse.cz>
3975         PR c/87483
3976         * cgraphunit.c (process_function_and_variable_attributes):
3977         Warn about a function with alias attribute and a body.
3979 2018-10-04  Martin Liska  <mliska@suse.cz>
3981         PR ipa/82625
3982         * multiple_target.c (redirect_to_specific_clone): New function.
3983         (ipa_target_clone): Use it.
3984         * tree-inline.c: Fix comment.
3986 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
3988         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
3989         fields.
3990         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
3991         (gcc::dump_manager::register_pass): New member function, adapted
3992         from loop body in gcc::pass_manager::register_pass, adding a
3993         call to update_dfi_for_opt_info.
3994         (gcc::dump_manager::opt_info_enable_passes): Store the
3995         -fopt-info options into the new fields.  Move the loop
3996         bodies into...
3997         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
3998         function.
3999         * dumpfile.h (struct opt_pass): New forward decl.
4000         (gcc::dump_manager::register_pass): New decl.
4001         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
4002         (class gcc::dump_manager): Add fields "m_optgroup_flags",
4003         "m_optinfo_flags", and "m_optinfo_filename".
4004         * passes.c (gcc::pass_manager::register_pass): Move all of the
4005         dump-handling code to gcc::dump_manager::register_pass.
4007 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
4009         PR rtl-optimization/87466
4010         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
4011         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
4012         * doc/tm.texi: Regenerate.
4013         * ira-lives.c (process_bb_node_lives): Use the new target hook.
4014         * lra-lives.c (process_bb_lives): Likewise.
4015         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
4016         Define.
4018 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
4020         * params.c (add_params): Fix initialization.
4022 2018-10-04  Martin Liska  <mliska@suse.cz>
4024         PR gcov-profile/84107
4025         * tree-profile.c (init_ic_make_global_vars):
4026         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
4027         Come up with new ic_tuple* variables.  Emit
4028         __gcov_indirect_call{,_topn} variables.
4029         (gimple_gen_ic_profiler): Access the variable
4030         and emit gimple.
4031         (gimple_gen_ic_func_profiler): Access
4032         __gcov_indirect_call.callee field.
4033         (gimple_init_gcov_profiler): Use ptr_type_node.
4034         * value-prof.c (gimple_ic): Use ptr_type_node.
4036 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
4038         PR tree-optimization/85787
4039         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
4040         into this function and add support for detecting multiple phis.
4041         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
4043 2018-10-04  Martin Liska  <mliska@suse.cz>
4045         PR ipa/87491
4046         * ipa-inline.c (inline_to_all_callers_1):
4047         Call ultimate_alias_target for node being inlined.
4049 2018-10-03  Jeff Law  <law@redhat.com>
4051         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
4052         target's wchar_t.
4053         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
4054         * tree.h (get_typenode_from_name): Prototype.
4056 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
4058         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
4059         Change operand 2 predicate to nonimmediate_operand.
4060         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4062 2018-10-03  Martin Sebor  <msebor@redhat.com>
4063             Jeff Law  <law@redhat.com>
4065         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
4066         initialize it.
4067         (get_string_length): Detect unterminated arrays.
4068         (format_string): Same.
4069         (format_directive): Warn about unterminated arrays.
4070         (handle_gimple_call): Mark statements with no_warning as needed.
4072 2018-10-03  Jim Wilson  <jimw@sifive.com>
4074         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
4075         also define __riscv_abi_rve.  Delete trailing white space.
4077 2018-10-03  Paul Koning  <ni1d@arrl.net>
4079         Enable LRA register allocator for PDP11.
4080         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
4081         (R): Likewise.
4082         (D): Likewise.
4083         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
4084         * config/pdp11/pdp11.opt (-mlra): New option.
4085         * doc/invoke.texi (PDP-11 Options): Document -mlra.
4087 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
4089         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
4090         (*<absneg:code>extend<mode>xf2): Ditto.
4092 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
4094         PR tree-optimization/87415
4095         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
4096         precision fields.
4098 2018-10-02  Jeff Law  <law@redhat.com>
4100         * gimple-fold.c (get_range_strlen): Only set *nonstr when
4101         an unterminated string is discovered.  Bubble up range
4102         even for unterminated strings.
4103         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
4104         indicates the string was not terminated via NONSTR.
4106 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
4108         * tree-vrp.c (extract_range_from_unary_expr): Special case all
4109         pointer conversions.
4110         Do not do anything special for anti-ranges.
4112 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
4114         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
4115         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
4117 2018-10-03  Martin Liska  <mliska@suse.cz>
4119         PR gcov-profile/86109
4120         * coverage.c (coverage_begin_function): Do not
4121         mark lambdas as artificial.
4122         * tree-core.h (struct GTY): Remove tm_clone_flag
4123         and introduce new lambda_function.
4124         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
4126 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
4128         PR target/87474
4129         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
4130         P8_VECTOR and VSX are enabled.
4132 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
4134         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
4135         0x3907 as CPU model number.
4137 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
4139         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
4140         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
4141         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
4142         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
4143         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
4144         * config/s390/s390.md: Likewise. Rename also the cpu attribute
4145         value from arch12 to z14.
4147 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
4149         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
4150         (isinfxf2): Ditto.
4151         (isinf<mode>2): Ditto.
4153 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
4155         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
4156         before emitting fxam.  Perform calculations in XFmode.
4158 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
4160         * match.pd (((X /[ex] A) +- B) * A): New transformation.
4162 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
4164         PR middle-end/87319
4165         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
4166         * tree.c (signed_or_unsigned_type_for): Handle complex.
4168 2018-10-02  Jeff Law  <law@redhat.com>
4170         * gimple-fold.c (get_range_strlen): Remove dead code.
4172 2018-10-02  Martin Sebor  <msebor@redhat.com>
4173             Jeff Law  <law@redhat.com>
4175         * builtins.c (unterminated_array): Add new arguments.
4176         If argument is not terminated, bubble up size and exact
4177         state to callers.
4178         (expand_builtin_strnlen): Detect, avoid expanding
4179         and diagnose unterminated arrays.
4180         (c_strlen): Fill in offset of start of unterminated strings.
4181         * builtins.h (unterminated_array): Update prototype.
4183 2018-10-02  Richard Biener  <rguenther@suse.de>
4185         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
4186         of haddv4df, first reduce to SSE width and exploit the fact
4187         that we only need element zero with the reduction result.
4188         (reduc_plus_scal_v2df): Likewise.
4190 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
4192         * dojump.h (do_jump): Delete.
4193         (do_jump_1): Likewise.
4194         (split_comparison): Move around.
4195         * dojump.c (do_jump): Make static.
4196         (do_jump_1): Likewise.
4197         (jumpifnot): Move around.
4198         (jumpifnot_1): Likewise.
4199         (jumpif): Likewise.
4200         (jumpif_1): Likewise.
4201         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
4203 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
4205         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
4206         insns in the delay slot and add_insn_after for the jump insn.
4208 2018-10-02  Richard Biener  <rguenther@suse.de>
4210         * tree-inline.c (expand_call_inline): Use the location of
4211         the callee declaration for the inline-entry marker.
4212         * final.c (notice_source_line): Remove special-casing of
4213         NOTE_INSN_INLINE_ENTRY.
4215 2018-10-01  Carl Love  <cel@us.ibm.com>
4217         PR 69431
4218         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
4219         (__builtin_mtfsb0): New.
4220         (__builtin_mtfsb1): New.
4221         ( __builtin_set_fpscr_rn): New.
4222         (__builtin_set_fpscr_drn): New.
4223         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
4224         (rs6000_expand_set_fpscr_rn_builtin): Add.
4225         (rs6000_expand_set_fpscr_drn_builtin): Add.
4226         (rs6000_expand_builtin): Add case statement entries for
4227         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
4228         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
4229         RS6000_BUILTIN_MFFSL.
4230         (rs6000_init_builtins): Add ftype initialization and def_builtin
4231         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
4232         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
4233         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
4234         rs6000_mffscdrn): Add define_insn.
4235         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
4236         * doc/extend.texi: Add documentation for the builtins.
4238 2018-10-01  Richard Biener  <rguenther@suse.de>
4240         PR tree-optimization/87465
4241         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
4242         causing branch miscounts.
4244 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4246         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
4247         aarch64_option_default_param):  New.
4248         (params.h): Include.
4249         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
4250         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
4251         stack-clash protection validation code.
4253 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4255         * params.c (validate_param): New.
4256         (add_params): Use it.
4257         (set_param_value): Refactor param validation into validate_param.
4258         (diagnostic.h): Include.
4259         * diagnostic.h (diagnostic_ready_p): New.
4261 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4263         * params.c (set_param_value):
4264         Add index of parameter being validated.
4265         * common/common-target.def (option_validate_param): New.
4266         * common/common-targhooks.h (default_option_validate_param): New.
4267         * common/common-targhooks.c (default_option_validate_param): New.
4268         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
4269         * doc/tm.texi: Regenerate.
4271 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4273         PR target/86486
4274         * config/aarch64/aarch64.c (aarch64_override_options_internal):
4275         Add validation for stack-clash parameters and set defaults.
4277 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4279         PR target/86486
4280         * configure.ac: Add stack-clash-protection-guard-size.
4281         * doc/install.texi: Document it.
4282         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
4283         * params.def: Update comment for guard-size.
4284         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
4285         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
4286         * configure: Regenerate.
4288 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4290         PR target/86486
4291         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
4292         STACK_DYNAMIC_OFFSET): New.
4293         * config/aarch64/aarch64.c (aarch64_layout_frame):
4294         Update outgoing args size.
4295         (aarch64_stack_clash_protection_alloca_probe_range,
4296         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
4298 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4300         PR target/86486
4301         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
4302         probe ranges.
4303         * target.def (stack_clash_protection_alloca_probe_range): New.
4304         (stack_clash_protection_final_dynamic_probe): Remove.
4305         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
4306         (default_stack_clash_protection_final_dynamic_probe): Remove.
4307         * targhooks.c: Likewise.
4308         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
4309         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
4310         * doc/tm.texi: Regenerate.
4312 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4314         PR target/86486
4315         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
4316         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
4317         aarch64_clamp_to_uimm12_shift): New.
4318         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
4319         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
4321 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
4323         PR target/86486
4324         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
4326 2018-10-01  Jeff Law  <law@redhat.com>
4327             Richard Sandiford <richard.sandiford@linaro.org>
4328             Tamar Christina  <tamar.christina@arm.com>
4330         PR target/86486
4331         * config/aarch64/aarch64.md
4332         (probe_stack_range): Add k (SP) constraint.
4333         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
4334         STACK_CLASH_MAX_UNROLL_PAGES): New.
4335         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
4336         stack probes for stack clash.
4337         (aarch64_allocate_and_probe_stack_space): New.
4338         (aarch64_expand_prologue): Use it.
4339         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
4340         (aarch64_sub_sp): Add emit_move_imm optional param.
4342 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
4344         PR tree-optimization/87261
4345         * match.pd: Remove trailing whitespace.
4346         Add (x & y) | ~(x | y) -> ~(x ^ y),
4347         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
4349 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
4351         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
4352         constraints.
4353         (ashlsi3_insn): Update instruction constraints.
4354         (ashrsi3_insn): Likewise.
4355         (rotrsi3): Likewise.
4356         (add_shift): Likewise.
4357         * config/arc/constraints.md (Csz): New 32 bit constraint. It
4358         avoids placing in the limm field small constants which, otherwise,
4359         could end into a small instruction.
4361 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
4363         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
4364         destination register is not odd-even.
4365         (umaddsidi4_split): Likewise.
4367 2018-10-01  Richard Biener  <rguenther@suse.de>
4369         * tree-inline.c (expand_call_inline): Store origin of fn
4370         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
4371         * tree.c (block_ultimate_origin): Simplify and do some
4372         checking.
4374 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
4376         * config/i386/mmx.md (EMMS): New int iterator.
4377         (emms): New int attribute.
4378         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
4379         EMMS int iterator.  Explicitly declare clobbers.
4380         (mmx_emms): Remove expander.
4381         (mmx_femms): Ditto.
4382         * config/i386/predicates.md (emms_operation): Remove predicate.
4383         (vzeroall_pattern): New predicate.
4384         (vzeroupper_pattern): Rename from vzeroupper_operation.
4385         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
4386         vzeroupper_pattern and vzeroall_pattern predicates.
4388 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
4390         PR rtl-optimization/86939
4391         * ira-lives.c (make_hard_regno_born): Rename from this...
4392         (make_hard_regno_live): ... to this.  Remove update to conflict
4393         information.  Update function comment.
4394         (make_hard_regno_dead): Add conflict information update.  Update
4395         function comment.
4396         (make_object_born): Rename from this...
4397         (make_object_live): ... to this.  Remove update to conflict information.
4398         Update function comment.
4399         (make_object_dead):  Add conflict information update.  Update function
4400         comment.
4401         (mark_pseudo_regno_live): Call make_object_live.
4402         (mark_pseudo_regno_subword_live): Likewise.
4403         (mark_hard_reg_dead): Update function comment.
4404         (mark_hard_reg_live): Call make_hard_regno_live.
4405         (process_bb_node_lives): Likewise.
4406         * lra-lives.c (make_hard_regno_born): Rename from this...
4407         (make_hard_regno_live): ... to this.  Remove update to conflict
4408         information.  Remove now uneeded check_pic_pseudo_p argument.
4409         Update function comment.
4410         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
4411         to conflict information.  Update function comment.
4412         (mark_pseudo_live): Remove update to conflict information.  Update
4413         function comment.
4414         (mark_pseudo_dead): Add conflict information update.
4415         (mark_regno_live): Call make_hard_regno_live.
4416         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
4417         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
4419 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
4421         PR target/87370
4422         * config/i386/i386.c (construct_container): Use TImode for
4423         BLKmode values in 2 integer registers.
4425 2018-09-29  Jeff Law  <law@redhat.com>
4427         * builtins.c (unterminated_array): Pass in c_strlen_data * to
4428         c_strlen rather than just a tree *.
4429         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
4430         Update recursive calls appropriately.  If caller did not provide a
4431         suitable data pointer, create a local one.  When a non-terminated
4432         string is discovered, bubble up information about the string via the
4433         c_strlen_data object.
4434         * builtins.h (c_strlen): Update prototype.
4435         (c_strlen_data): New structure.
4436         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
4437         For a type 2 call, if c_strlen indicates a non-terminated string
4438         use the length of the non-terminated string.
4439         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
4441 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
4443         PR target/87467
4444         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
4445         __m512d type for __A argument rather than __m512.
4447 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
4449         * match.pd (simple_comparison): Don't optimize if either operand is
4450         a function pointer when target needs function pointer canonicalization.
4452 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
4454         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
4455         power5 .. power9 to remove indirection.
4456         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
4457         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
4458         ASM_CPU_476_SPEC): Delete.
4459         (ASM_CPU_SPEC): Adjust.
4460         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
4461         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
4463 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
4465         * config.in: Delete HAVE_AS_DCI.
4466         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
4467         * config/rs6000/rs6000.h: Ditto.
4468         * configure.ac: Delete HAVE_AS_DCI.
4469         * configure: Regenerate.
4471 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
4473         * config.in (HAVE_AS_LWSYNC): Delete.
4474         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
4475         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
4476         do it as a .long .
4477         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
4478         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
4479         as a .long .
4480         * configure.ac: Delete HAVE_AS_LWSYNC.
4481         * configure: Regenerate.
4483 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
4484             Pierre-Marie de Rodat  <derodat@adacore.com>
4486         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
4487         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
4488         (cgraph_node::create_thunk): Add indirect_offset parameter.
4489         (thunk_adjust): Likewise.
4490         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
4491         and initialize the corresponding field with it.
4492         (cgraph_node::dump): Dump indirect_offset field.
4493         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
4494         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
4495         (thunk_adjust): Add indirect_offset parameter and deal with it.
4496         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
4497         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
4498         or if the thunk is external or local.  Fix formatting.  Do not chain
4499         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
4500         if any, in the GIMPLE representation.
4501         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
4502         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
4503         (input_node): Read indirect_offset field.
4504         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
4505         call to thunk_adjust.
4506         * tree-nested.c (struct nesting_info): Add thunk_p field.
4507         (create_nesting_tree): Set it.
4508         (convert_all_function_calls): Copy static chain from targets to thunks.
4509         (finalize_nesting_tree_1): Return early for thunks.
4510         (unnest_nesting_tree_1): Do not finalize thunks.
4511         (gimplify_all_functions): Do not gimplify thunks.
4513 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
4515         * opt-suggestions.c (option_proposer::build_option_suggestions):
4516         Release "option_values".
4518 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
4520         * coverage.c (get_coverage_counts): Convert problem-reporting dump
4521         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
4522         * dumpfile.c (kind_as_string): New function.
4523         (dump_loc): Rather than a hardcoded prefix of "note: ", use
4524         kind_as_string to vary the prefix based on dump_kind.
4525         (selftest::test_capture_of_dump_calls): Update for above.
4527 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
4529         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
4530         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
4532 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
4534         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
4535         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
4536         INVALID_REGNUM instead of FPSR_REG.
4537         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
4538         * config/i386/i386.md: Update comment of FP compares.
4539         (fldenv): Do not clobber FPSR_REG.
4541 2018-09-28  Richard Biener  <rguenther@suse.de>
4543         * tree.h (BLOCK_ORIGIN): New.
4544         * omp-expand.c (grid_expand_target_grid_body): Assign
4545         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
4546         * tree-inline.c (remap_block): Likewise.
4547         * auto-profile.c (get_function_decl_from_block): Simplify
4548         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
4549         * langhooks.c (lhd_print_error_function): Likewise.
4550         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
4551         Likewise.
4552         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4553         * tree.c (block_nonartificial_location): Likewise.
4554         (block_ultimate_origin): Likewise.
4555         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
4556         no longer needed LTO case.
4558 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
4559             Jan Hubicka  <jh@suse.cz>
4560             Martin Jambor  <mjambor@suse.cz>
4562         * simplify-rtx.c (simplify_merge_mask): New function.
4563         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
4564         same masks are used in op1 or op2.
4565         (test_vec_merge): New function.
4566         (test_vector_ops): Call test_vec_merge.
4568 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
4570         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
4571         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
4572         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
4573         (cypress_costs): Set it.
4574         (supersparc_costs): Likewise.
4575         (hypersparc_costs): Likewise.
4576         (leon_cost): Likewise.
4577         (leon3_costs): Likewise.
4578         (sparclet_costs): Likewise.
4579         (ultrasparc_costs): Likewise.
4580         (ultrasparc_costs): Likewise.
4581         (niagara_costs): Likewise.
4582         (niagara2_costs): Likewise.
4583         (niagara3_costs): Likewise.
4584         (niagara4_costs): Likewise.
4585         (niagara7_costs): Likewise.
4586         (m8_costs): Likewise.
4587         (TARGET_CAN_FOLLOW_JUMP): Define.
4588         (pass_work_around_errata::gate): Minor tweak.
4589         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
4590         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
4591         Automaitcally clear MASK_FSMULD mask for V7 processors.
4592         (sparc_can_follow_jump): New static function.
4593         (output_ubranch): Deal with CROSSING_JUMP_P.
4594         (sparc_use_sched_lookahead): Rewrite using switch statement.
4595         (sparc_issue_rate): Reorder.
4596         (sparc_branch_cost): New function.
4598 2018-09-27  Martin Sebor  <msebor@redhat.com>
4600         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
4601         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
4602         (int_fits_type_p): Same.
4604 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
4606         * config/i386/i386.md (FPCR_REG): Remove.
4607         (UNSPEC_FLDCW): Remove.
4608         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
4609         (x86_fldcw_1): Remove insn pattern.
4610         (fnstenv): Do not clobber FPCR_REG.
4611         (fldenv): Ditto.
4612         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
4613         (CALL_USED_REGISTERS): Ditto.
4614         (REG_ALLOC_ORDER): Ditto.
4615         (REG_CLASS_CONTENTS): Ditto.
4616         (HI_REGISTER_NAMES): Ditto.
4617         (ADDITIONAL_REGISTER_NAMES): Use defines instead
4618         of numerical constants.
4619         * config/i386/i386.c (regclass_map): Remove fpsr register.
4620         (dbx_register_map): Ditto.
4621         (dbx64_register_map): Ditto.
4622         (svr4_dbx_register_map): Ditto.
4623         (print_reg): Do not handle FPCR_REG.
4625 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
4627         PR target/87149
4628         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
4629         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
4630         Delete, always treat as true.
4631         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
4632         Ditto.  Simplify remaining code.
4633         * config/powerpcspe/powerpcspe.h: Ditto.
4634         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
4635         Simplify remaining code.
4636         (rs6000_expand_builtin): Ditto.
4637         * config/rs6000/rs6000.h: Ditto.
4638         * configure.ac: Ditto.
4639         * configure: Regenerate.
4641 2018-09-27  Martin Liska  <mliska@suse.cz>
4643         * coverage.c (get_coverage_counts): Revert the formatting
4644         of missing profile opt info.
4646 2018-09-27  Richard Biener  <rguenther@suse.de>
4648         PR debug/37801
4649         PR debug/87440
4650         * dwarf2out.c (set_block_origin_self): Do not mark outermost
4651         block as we do not output that.
4652         (gen_inlined_subroutine_die): Elide the originally outermost
4653         block, matching what we do for concrete instances.
4654         (decls_for_scope): Add parameter specifying whether to recurse
4655         to subblocks.
4657 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
4658             Tom de Vries  <tom@codesourcery.com>
4660         PR 82089
4662         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
4663         STORE_FLAG_VALUE == 1.
4665 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
4667         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
4668         constant definitions.
4669         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
4670         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
4671         ("speculation_barrier"): New expander definition.
4673 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
4675         PR gcov-profile/86957
4676         * common.opt: New warning option -Wmissing-profile.
4677         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
4678         * doc/invoke.texi: Document -Wmissing-profile.
4680 2018-09-26  Jim Wilson  <jimw@sifive.com>
4682         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
4683         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
4685 2018-09-26  Martin Sebor  <msebor@redhat.com>
4687         * tree.c (zerop): Change return type to bool.
4688         (integer_zerop, integer_onep, integer_each_onep): Same.
4689         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
4690         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
4691         (real_onep, real_minus_onep, chain_index): Same.
4692         (print_type_hash_statistics, type_list_equal): Same.
4693         * tree.h (zerop): Same.
4694         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
4695         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
4696         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
4697         (real_onep, real_minus_onep, chain_index): Same.
4698         (print_type_hash_statistics, type_list_equal): Same.
4700 2018-09-26  Jim Wilson  <jimw@sifive.com>
4702         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
4704 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
4706         PR target/87414
4707         * config/i386/i386.c: Include debug.h and dwarf2out.h.
4708         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
4709         call.
4711 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
4713         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
4715 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
4717         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
4718         and FP_SECOND_SSE_REGS.
4719         (REG_CLASS_NAMES): Ditto.
4720         (REG_CLASS_CONTENTS): Ditto.
4721         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
4722         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
4723         (ix86_preferred_output_reload_class): Ditto.
4724         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
4725         clobber constraint to "=&f".
4726         (fix_truncdi_i387): Ditto.
4727         (lrintxfdi2): Ditto.
4728         (fistdi2_<rounding>): Ditto.
4729         (fpremxf4_i387): Change "=u" constraint to "=f".
4730         (fprem1xf4_i387): Ditto.
4731         (sincosxf3): Ditto.
4732         (fptanxf4_i387): Ditto.
4733         (fxtractxf3_i387): Ditto.
4734         (fscalexf4_i387): Ditto.
4735         (atan2xf3): Change "u" constraint to "f".
4736         (fyl2xxf3_i387): Ditto.
4737         (fyl2xp1xf3_i387): Ditto.
4739 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
4741         PR target/87439
4742         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
4743         for removed I387_MASK_PM entity.
4746 2018-09-26  Jeff Law  <law@redhat.com>
4747         Revert
4748         2018-09-26  Alexey Neyman  <stilor@att.net>
4750         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
4751         headers are no longer pulled in by <isl/val.h>.
4753 2018-09-26  Richard Biener  <rguenther@suse.de>
4755         PR debug/87443
4756         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
4757         or concrete instance DIE to the tree.  Create abstract origin
4758         attributes also for concrete instances.
4760 2018-09-26  Alexey Neyman  <stilor@att.net>
4762         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
4763         headers are no longer pulled in by <isl/val.h>.
4765 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
4767         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
4768         Use new helper functions.
4769         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
4770         Use new helper functions.
4771         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
4772         aarch_mm_needs_release): New declarations.
4773         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
4774         aarch_mm_needs_release): New.
4776 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
4778         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
4779         (arm32_output_mi_thunk): Deal with long calls.
4781 2018-09-26  Richard Biener  <rguenther@suse.de>
4783         PR debug/87428
4784         PR debug/87362
4785         * tree-inline.c (expand_call_inline): When the location
4786         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
4787         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
4788         the inserted BLOCK to make inlined_function_outer_scope_p
4789         recognize it.
4790         * dwarf2out.c (add_call_src_coords_attributes): Do not add
4791         coords for reserved locations.
4793 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
4795         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
4796         (*call_indirect_nonlocal_sysv<mode>): Ditto.
4797         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
4798         (*sibcall_nonlocal_sysv<mode>): Ditto.
4799         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
4800         (<bd>_<mode>): Ditto.
4801         (<bd>tf_<mode>): Ditto.
4803 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
4805         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
4806         control string as a list of templates instead of as C code.
4807         (*altivec_movti): Ditto.
4808         * config/rs6000/darwin.md (movdf_low_di): Ditto.
4810 2018-09-25  Jim Wilson  <jimw@sifive.com>
4812         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
4813         when target symbol is weak.
4815 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4817         PR c/87387
4818         * builtins.c (unterminated_array): Simplify.
4819         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
4820         where pointer arithmetic is safe.
4822 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
4824         PR target/86987
4825         * config/rs6000/altivec.md (altivec_vspltb): Use
4826         const_0_to_15_operand instead of u5bit_cint_operand.
4827         (*altivec_vspltb_internal): Ditto.
4828         (altivec_vspltb_direct): Ditto.
4829         (altivec_vsplth): Use const_0_to_7_operand instead of
4830         u5bit_cint_operand.
4831         (*altivec_vsplth_internal): Ditto.
4832         (altivec_vsplth_direct): Ditto.
4833         (altivec_vspltw): Use const_0_to_3_operand instead of
4834         u5bit_cint_operand.
4835         (*altivec_vspltw_internal): Ditto.
4836         (altivec_vspltw_direct): Ditto.
4837         (altivec_vspltsf): Ditto.
4838         (*altivec_vspltsf_internal): Ditto.
4839         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
4840         various splats with the proper size immediate.  Reorder the various
4841         cases by ascending size of immediate, and put all such together.
4843 2018-09-25  Richard Biener  <rguenther@suse.de>
4845         PR debug/83941
4846         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
4847         GC-ification.
4848         (maybe_create_die_with_external_ref): Do not create
4849         DW_TAG_imported_unit here.
4850         (add_abstract_origin_attribute): Handle external BLOCK refs.
4851         (dwarf2out_abstract_function): Simplify LTO case.
4852         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
4853         rather than using maybe_create_die_with_external_ref.
4855 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
4857         PR target/71278
4858         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
4859         (frndintxf2_mask_pm_i387): Ditto.
4860         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
4861         Enable for !flag_trapping_math.
4862         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
4863         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
4864         Change operand 1 predicate to nonimmediate_operand.
4865         (attr "i387_cw"): Remove mask_pm.
4866         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
4867         (enum ix86_entity): Remove I387_MASK_PM.
4868         * config/i386/i386.c (ix86_i387_mode_needed): Do not
4869         handle I387_MASK_PM.
4870         (ix86_mode_needed): Ditto.
4871         (ix86_mode_after): Ditto.
4872         (ix86_mode_entry): Ditto.
4873         (ix86_mode_exit): Ditto.
4874         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
4876 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
4878         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
4879         to_update_switch_stmts to vNULL instead of calling create on them
4880         immediately.
4882 2018-09-25  Richard Biener  <rguenther@suse.de>
4884         PR tree-optimization/87402
4885         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
4886         (visit_phi): Re-instantiate handling of supposed to be VARYING
4887         but non-VARYING backedge value.
4889 2018-09-25  Richard Biener  <rguenther@suse.de>
4891         PR debug/83941
4892         * dwarf2out.c (struct sym_off_pair): New.
4893         (external_die_map): New global.
4894         (lookup_decl_die): When in LTO create DIEs lazily from the
4895         external_die_map.
4896         (lookup_block_die): New function, create DIEs lazily in LTO.
4897         (equate_block_to_die): New function.
4898         (dwarf2out_die_ref_for_decl): During WPA get the association
4899         from the external DIE map.
4900         (dwarf2out_register_external_die): Record mapping into the
4901         external DIE map.
4902         (maybe_create_die_with_external_ref): New function split out from
4903         DIE generation part of old dwarf2out_register_external_die.
4904         (add_abstract_origin_attribute): Do not return the DIE.  When
4905         in LTO reference externals directly.
4906         (dwarf2out_abstract_function): When in LTO ignore calls for
4907         decls with external DIEs (already present abstract instances).
4908         (gen_call_site_die): Adjust.
4909         (add_high_low_attributes): Likewise.
4910         (gen_lexical_block_die): Likewise.
4911         (gen_inlined_subroutine_die): Likewie.
4912         (gen_block_die): Likewise.
4913         (dwarf2out_inline_entry): Likewise.
4914         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
4915         DIEs.
4917 2018-09-25  Martin Liska  <mliska@suse.cz>
4919         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
4920         integers and not by a float value.
4922 2018-09-25  Martin Liska  <mliska@suse.cz>
4924         PR fortran/87394
4925         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
4926         instead of NULL.
4927         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
4928         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
4929         * godump.c (go_format_type): Remove extra parenthesis.
4931 2018-09-25  Martin Liska  <mliska@suse.cz>
4933         * alias.c (set_dest_equal_p): Remove unused function.
4934         * config/i386/i386.c (def_builtin_pure2): Likewise.
4935         * diagnostic-show-locus.c (class layout): Remove
4936         unused field.
4937         (layout::layout): Likewise here.
4938         * dump-context.h (class temp_dump_context): Likewise.
4939         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
4940         (add_AT_loclistsptr): Likewise.
4941         (add_AT_offset): Likewise.
4942         (get_AT_hi_pc): Likewise.
4943         (is_comdat_die): Likewise.
4944         (type_is_enum): Likewise.
4945         (ceiling): Likewise.
4946         (add_AT_vms_delta): Likewise.
4947         (is_class_die): Likewise.
4948         * edit-context.c (class line_event): Remove unused field.
4949         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
4950         unused function.
4951         * ipa-cp.c (ipa_get_vr_lat): Likewise.
4952         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
4953         (ok_for_base_p_nonstrict): Likewise.
4954         * tree-chrec.c (is_not_constant_evolution): Likewise.
4955         (chrec_fold_poly_cst): Likewise.
4956         * tree-if-conv.c (has_pred_critical_p): Likewise.
4957         * tree-ssa-coalesce.c (print_exprs): Likewise.
4958         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
4959         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
4960         * tree-vrp.c (value_ranges_intersect_p): Likewise.
4961         (value_range_nonnegative_p): Likewise.
4963 2018-09-25  Martin Liska  <mliska@suse.cz>
4965         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
4966         Do not handle "GNU Pascal".
4967         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
4968         Likewise.
4969         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
4970         from documentation. Likewise.
4971         * dbxout.c (dbxout_range_type): Likewise.
4972         * doc/cpp.texi: Likewise.
4973         * doc/extend.texi: Likewise.
4974         * doc/frontends.texi: Likewise.
4975         * doc/invoke.texi: Remove Pascal entry.
4976         * tree.def (CLEANUP_POINT_EXPR): Likewise.
4977         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
4979 2018-09-25  Martin Liska  <mliska@suse.cz>
4981         PR middle-end/86078
4982         * doc/invoke.texi: Document all parameters and remove default
4983         of the parameters.
4985 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
4987         PR bootstrap/87417
4988         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
4989         contains HOST_WIDE_INTs when computing its size.
4991 2018-09-24  Jim Wilson  <jimw@sifive.com>
4993         PR target/87391
4994         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
4995         not TARGET_RVE.
4996         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
4998 2018-09-24  Andrew Pinski  <apinski@marvell.com>
5000         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
5001         access prev before checking it for NULLness in the
5002         AARCH64_FUSE_CMP_BRANCH case.
5004 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
5006         PR target/82699
5007         * config/i386/i386.c (rest_of_insert_endbranch): Set
5008         endbr_queued_at_entrance to true and don't insert ENDBR if
5009         x86_function_profiler will be called.
5010         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
5011         is true.
5012         * config/i386/i386.h (machine_function): Add
5013         endbr_queued_at_entrance.
5015 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
5017         * genattrtab.c (mk_attr_alt): Use alternative_mask.
5018         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
5019         types.
5020         (check_attr_test): Use alternative_mask.
5021         (get_attr_value): Likewise.
5022         (compute_alternative_mask): Use alternative_mask and XWINT.
5023         (make_alternative_compare): Use alternative_mask.
5024         (attr_alt_subset_p): Use XWINT.
5025         (attr_alt_subset_of_compl_p): Likewise.
5026         (attr_alt_intersection): Use alternative_mask and XWINT.
5027         (attr_alt_union): Likewise.
5028         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
5029         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
5030         (simplify_test_exp): Use alternative_mask and XWINT.
5031         (write_test_expr): Use alternative_mask and XWINT, adjust bit
5032         number calculation to support 64 bits.  Generate code that
5033         checks 64-bit masks.
5034         (main): Use alternative_mask.
5035         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
5037 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
5039         PR target/80080
5040         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
5041         RETURN+USE when returning via %r14.
5043 2018-09-24  Martin Liska  <mliska@suse.cz>
5045         * gcov.c (output_lines): Print colorization legend
5046         for both flag_use_colors and flag_use_hotness_colors.
5047         Reword the help.
5049 2018-09-24  Martin Liska  <mliska@suse.cz>
5051         * coverage.c (get_coverage_counts): Use warning_at
5052         with current_function_decl location. Use %qD in warning
5053         message.
5055 2018-09-24  Martin Liska  <mliska@suse.cz>
5057         * memory-block.h (memory_block_pool::release): Annotate with
5058         valgrind that the memory is not accessible.
5060 2018-09-24  Martin Liska  <mliska@suse.cz>
5062         PR sanitizer/85774
5063         * asan.c: Make asan_handled_variables extern.
5064         * asan.h: Likewise.
5065         * cfgexpand.c (expand_stack_vars): Make sure
5066         a representative is unpoison if another
5067         variable in the partition is handled by
5068         use-after-scope sanitization.
5070 2018-09-24  Richard Biener  <rguenther@suse.de>
5072         PR tree-optimization/63155
5073         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
5074         the worklist when the edge of the respective argument isn't
5075         executable.
5077 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
5079         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
5080         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
5081         (MASK_CLASS_P): Update for rename.
5082         (MAYBE_MASK_CLASS_P): Ditto.
5083         (REG_CLASS_NAMES): Update.
5084         (REG_CLASS_CONTENT): Update.
5085         * config/i386/i386.c (regclass_map): Update for MASK_REG
5086         and ALL_MASK_REGS rename.
5087         * config/i386/constraints.md (Yk): Update for rename.
5088         (k): Ditto.
5090 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
5092         * config/i386/i386.h (enum reg_class): Remove
5093         EVEX_SSE_REGS and MOD4_SSE_REGS.
5094         (REG_CLASS_NAMES): Update.
5095         (REG_CLASS_CONTENT): Update.
5096         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
5097         registers as ALL_SSE_REGS.
5098         (ix86_additional_allocno_class_p): Remove.
5099         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
5100         (ix86_register_priority): Lower priority of EVEX SSE registers.
5101         Use IN_RANGE macro where appropriate.
5102         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
5103         AVX-5124VNNIW checks.
5104         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
5105         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
5106         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
5107         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
5108         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
5109         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
5110         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
5111         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
5112         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
5113         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
5114         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
5115         * config/i386/constraints.md (Yh): Remove.
5117 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
5119         * config/i386/i386.c (regclass_map): Declare integer REX registers
5120         as GENERAL_REGS.
5122 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
5124         * doc/service.texi (Service): Switch the fsf.org link to https.
5126 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
5128         PR target/86798
5129         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5130         Define to speculation_safe_value_not_needed.
5132 2018-09-21  Florian Weimer  <fweimer@redhat.com>
5134         PR middle-end/81035
5135         * doc/extend.texi (Common Function Attributes): Mention that
5136         noreturn suppresses tail call optimization.
5138 2018-09-21  Jeff Law  <law@redhat.com>
5140         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
5141         vr_values::cleanup_edges_and_switches.
5142         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
5143         vr_values class.
5144         (identify_jump_threads): Remove EDGE_IGNORE handling.
5145         (execute_vrp): Move handling of to_remove_edges and
5146         to_update_switch_stmts into vr_values class member functions.
5147         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
5148         (to_update_switch_stmts): Likewise.
5149         * vr-values.c: Include cfghooks.h.
5150         (vr_values::vr_values): Initialize to_remove_edges and
5151         to_update_switch_stmts.
5152         (vr_values::~vr_values): Verify to_remove_edges and
5153         to_update_switch_stmts are empty.
5154         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
5155         (vr_values::cleanup_edges_and_switches): New member function.
5156         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
5157         function.  Add new data members.
5159 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
5161         PR tree-optimization/87309
5162         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
5163         calls with pflags and alt_flags.
5164         (selftest::test_capture_of_dump_calls): Add test of interaction of
5165         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
5167 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5169         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
5171 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5173         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
5175 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5177         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
5178         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
5179         (VXWORKS_LIBS_RTP): Minor reordering.
5181 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5183         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
5184         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
5186 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5188         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
5189         (PTRDIFF_TYPE): Likewise.
5191 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5193         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
5194         triplet, similar to support for VxWorks7.
5195         * config/vxworks-dummy.h: Provide a default definition
5196         of TARGET_VXWORKS64 to 0.
5198 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5200         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
5201         * config/vxworks-dummy.h: here.
5203 2018-09-21  Olivier Hainque  <hainque@adacore.com>
5205         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
5207 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
5208             Bo Zhou  <zbo.zhou@hisilicon.com>
5210         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
5211         * config/aarch64/aarch64-tune.md: Regenerated.
5212         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
5213         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
5214         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
5216 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
5217             Julian Brown  <julian@codesourcery.com>
5219         * builtins.c (get_builtin_sync_mem): Handle address spaces.
5221 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
5223         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
5224         if the call takes a static chain.
5226 2018-09-21  Martin Liska  <mliska@suse.cz>
5228         * auto-profile.c (autofdo_source_profile::read): Do not
5229         set sum_all.
5230         (read_profile): Do not add working sets.
5231         (read_autofdo_file): Remove sum_all.
5232         (afdo_callsite_hot_enough_for_early_inline): Remove const
5233         qualifier.
5234         * coverage.c (struct counts_entry): Remove gcov_summary.
5235         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
5236         do not support GCOV_TAG_PROGRAM_SUMMARY.
5237         (get_coverage_counts): Remove summary and expected
5238         arguments.
5239         * coverage.h (get_coverage_counts): Likewise.
5240         * doc/gcov-dump.texi: Remove -w option.
5241         * gcov-dump.c (dump_working_sets): Remove.
5242         (main): Do not support '-w' option.
5243         (print_usage): Likewise.
5244         (tag_summary): Likewise.
5245         * gcov-io.c (gcov_write_summary): Do not dump
5246         histogram.
5247         (gcov_read_summary): Likewise.
5248         (gcov_histo_index): Remove.
5249         (gcov_histogram_merge): Likewise.
5250         (compute_working_sets): Likewise.
5251         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
5252         it not obsolete.
5253         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
5254         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
5255         (GCOV_HISTOGRAM_SIZE): Remove.
5256         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
5257         (struct gcov_summary): Simplify rapidly just
5258         to runs and sum_max fields.
5259         (gcov_histo_index): Remove.
5260         (NUM_GCOV_WORKING_SETS): Likewise.
5261         (compute_working_sets): Likewise.
5262         * gcov-tool.c (print_overlap_usage_message): Remove
5263         trailing empty line.
5264         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
5265         (output_lines): Remove program related line.
5266         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
5267         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
5268         histogram.
5269         (input_profile_summary): Do not read it.
5270         (merge_profile_summaries): And do not merge it.
5271         (input_symtab): Do not call removed function.
5272         * modulo-sched.c (sms_schedule): Do not print sum_max.
5273         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
5274         removed when histogram method was invented.
5275         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
5276         mode.
5277         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
5278         GCOV coding style.
5279         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
5280         and dump selected value.
5281         * profile.c (add_working_set): Remove.
5282         (get_working_sets): Likewise.
5283         (find_working_set): Likewise.
5284         (get_exec_counts): Do not work with working sets.
5285         (read_profile_edge_counts): Do not inform as sum_max is removed.
5286         (compute_branch_probabilities): Likewise.
5287         (compute_value_histograms): Remove argument for call of
5288         get_coverage_counts.
5289         * profile.h: Do not make gcov_summary const.
5291 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
5293         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
5295 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
5297         PR tree-optimization/86990
5298         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
5299         Check that the entire merged store group is made of constants only for
5300         overlapping stores.
5302 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
5304         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
5305         (VTABLE_VERIFICATION_SPEC): Likewise.
5306         (SANITIZER_EARLY_SPEC): Likewise.
5307         (SANITIZER_SPEC): Likewise.
5308         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
5309         * doc/invoke.texi (Link Options): Document -r.
5311 2018-09-20  Richard Biener <rguenther@suse.de>
5313         PR middle-end/87054
5314         * gimplify.c (gimplify_expr): Retain alignment of
5315         addressable lvalue in dereference.
5317 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
5319         PR bootstrap/87013
5320         * configure.ac: Check for .loc is_stmt support.
5321         * configure, config.in: Rebuilt.
5322         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
5323         if not supported.
5325 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
5327         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
5328         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
5329         -misel=no.
5331 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
5333         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5334         VECTOR_OTHER.
5335         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
5336         case VECTOR_OTHER.
5338 2018-09-20  Marek Polacek  <polacek@redhat.com>
5340         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
5342 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
5344         PR tree-optimization/87288
5345         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
5346         into account when determining PEELING_FOR_NITERS.
5348 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
5350         PR tree-optimization/86877
5351         * tree-vect-loop.c (vect_analyze_loop_2): Call
5352         vect_verify_datarefs_alignment.
5354 2018-09-19  Marek Polacek  <polacek@redhat.com>
5356         * doc/invoke.texi: Document -Wclass-conversion.
5358 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
5360         * config/pa/pa.c (pa_adjust_priority): Delete.
5361         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
5363         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
5364         (atomic_storehi): Likewise.
5365         (atomic_storesi): Likewise.
5366         (atomic_loaddi): Restore compare and swap exchange loop code.
5368 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
5370         PR rtl-optimization/86902
5371         * combine.c (try_combine): When changing the CC mode used, don't change
5372         an unrelated mode in other_insn to that new CC mode.
5374 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
5376         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
5377         with %T in place of calls to dump_generic_expr.
5378         (prune_runtime_alias_test_list): Likewise.
5379         (create_runtime_alias_checks): Likewise.
5380         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
5381         (vect_analyze_data_ref_dependence): Likewise.
5382         (vect_slp_analyze_data_ref_dependence): Likewise.
5383         (vect_record_base_alignment): Likewise.  Use %G in place of call
5384         to dump_gimple_stmt.
5385         (vect_compute_data_ref_alignment): Likewise.
5386         (verify_data_ref_alignment): Likewise.
5387         (vect_find_same_alignment_drs): Likewise.
5388         (vect_analyze_group_access_1): Likewise.
5389         (vect_analyze_data_ref_accesses): Likewise.
5390         (dependence_distance_ge_vf): Likewise.
5391         (dump_lower_bound): Likewise.
5392         (vect_prune_runtime_alias_test_list): Likewise.
5393         (vect_find_stmt_data_reference): Likewise.
5394         (vect_analyze_data_refs): Likewise.
5395         (vect_create_addr_base_for_vector_ref): Likewise.
5396         (vect_create_data_ref_ptr): Likewise.
5397         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
5398         (vect_can_advance_ivs_p): Likewise.
5399         (vect_update_ivs_after_vectorizer): Likewise.
5400         (vect_gen_prolog_loop_niters): Likewise.
5401         (vect_prepare_for_masked_peels): Likewise.
5402         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
5403         (vect_determine_vectorization_factor): Likewise.
5404         (vect_is_simple_iv_evolution): Likewise.
5405         (vect_analyze_scalar_cycles_1): Likewise.
5406         (vect_analyze_loop_operations): Likewise.
5407         (report_vect_op): Likewise.
5408         (vect_is_slp_reduction): Likewise.
5409         (check_reduction_path): Likewise.
5410         (vect_is_simple_reduction): Likewise.
5411         (vect_create_epilog_for_reduction): Likewise.
5412         (vect_finalize_reduction:): Likewise.
5413         (vectorizable_induction): Likewise.
5414         (vect_transform_loop_stmt): Likewise.
5415         (vect_transform_loop): Likewise.
5416         (optimize_mask_stores): Likewise.
5417         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
5418         (vect_split_statement): Likewise.
5419         (vect_recog_over_widening_pattern): Likewise.
5420         (vect_recog_average_pattern): Likewise.
5421         (vect_determine_min_output_precision_1): Likewise.
5422         (vect_determine_precisions_from_range): Likewise.
5423         (vect_determine_precisions_from_users): Likewise.
5424         (vect_mark_pattern_stmts): Likewise.
5425         (vect_pattern_recog_1): Likewise.
5426         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5427         (vect_record_max_nunits): Likewise.
5428         (vect_build_slp_tree_1): Likewise.
5429         (vect_build_slp_tree_2): Likewise.
5430         (vect_print_slp_tree): Likewise.
5431         (vect_analyze_slp_instance): Likewise.
5432         (vect_detect_hybrid_slp_stmts): Likewise.
5433         (vect_detect_hybrid_slp_1): Likewise.
5434         (vect_slp_analyze_operations): Likewise.
5435         (vect_slp_analyze_bb_1): Likewise.
5436         (vect_transform_slp_perm_load): Likewise.
5437         (vect_schedule_slp_instance): Likewise.
5438         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
5439         (vect_mark_stmts_to_be_vectorized): Likewise.
5440         (vect_init_vector_1): Likewise.
5441         (vect_get_vec_def_for_operand): Likewise.
5442         (vect_finish_stmt_generation_1): Likewise.
5443         (vect_check_load_store_mask): Likewise.
5444         (vectorizable_call): Likewise.
5445         (vectorizable_conversion): Likewise.
5446         (vectorizable_operation): Likewise.
5447         (vectorizable_load): Likewise.
5448         (vect_analyze_stmt): Likewise.
5449         (vect_is_simple_use): Likewise.
5450         (vect_get_vector_types_for_stmt): Likewise.
5451         (vect_get_mask_type_for_stmt): Likewise.
5452         * tree-vectorizer.c (increase_alignment): Likewise.
5454 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
5456         * doc/rtl.texi: Adjust vec_select description.
5457         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
5458         non-constant selectors.
5460 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
5462         * config/aarch64/aarch64-protos.h
5463         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
5464         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
5465         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
5466         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
5467         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
5468         (AARCH64_FL_PROFILE): Move index so flags are ordered.
5469         (AARCH64_ISA_RCPC8_4): New flag.
5470         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
5471         to aarch64_offset_9bit_signed_unscaled_p.
5472         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
5473         and use stlur.
5474         * config/aarch64/constraints.md (Ust): New constraint.
5475         * config/aarch64/predicates.md.
5476         (aarch64_9bit_offset_memory_operand): New predicate.
5477         (aarch64_rcpc_memory_operand): New predicate.
5479 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
5481         PR rtl-optimization/87361
5482         * rtlanal.c (nonzero_bits1): Revert accidental change.
5484 2018-09-19  Richard Biener  <rguenther@suse.de>
5486         PR tree-optimization/87349
5487         PR tree-optimization/87342
5488         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
5490 2018-09-18  Marek Polacek  <polacek@redhat.com>
5492         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
5493         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
5495 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
5497         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
5499 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
5501         PR rtl-optimization/86882
5502         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
5504 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
5506         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
5507         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
5509 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
5511         PR other/87353
5512         * doc/invoke.texi (Link Options): Fix formatting and grammar.
5514 2018-09-18  Richard Biener  <rguenther@suse.de>
5516         PR middle-end/63155
5517         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
5518         (compute_samebase_partition_bases): Likewise.
5519         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
5520         (gimple_can_coalesce_p): Simplify.
5522 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
5524         Handle a library implementation of ffs calling __builtin_ffs.
5525         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
5526         (mmix_init_libfuncs): New function: make __builtin_ffs expand
5527         to __ffsdi2.
5529 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
5531         * diagnostic-show-locus.c (class layout_range): Add field
5532         "m_original_idx".
5533         (layout_range::layout_range): Add "original_idx" param and use it
5534         to initialize new field.
5535         (make_range): Use 0 for original_idx.
5536         (layout::layout): Pass in index to calls to
5537         maybe_add_location_range.
5538         (layout::maybe_add_location_range): Add param "original_idx" and
5539         pass it on to layout_range.
5540         (layout::print_any_labels): Pass on range->m_original_idx to
5541         get_text call.
5542         (gcc_rich_location::add_location_if_nearby): Use 0 for
5543         original_idx.
5544         * gcc-rich-location.h (text_range_label::get_text): Update for new
5545         param.
5546         (range_label_for_type_mismatch::get_text): Likewise.
5548 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
5550         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
5552 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
5554         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
5555         format_string_diagnostic_t.
5556         (fmtwarn_n): Likewise.
5557         * substring-locations.c
5558         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
5559         (format_warning_n_va): Convert to...
5560         (format_string_diagnostic_t::emit_warning_n_va): ...this.
5561         (format_warning_va): Convert to...
5562         (format_string_diagnostic_t::emit_warning_va): ...this.
5563         (format_warning_at_substring): Convert to...
5564         (format_string_diagnostic_t::emit_warning): ...this.
5565         (format_warning_at_substring_n): Convert to...
5566         (format_string_diagnostic_t::emit_warning_n): ...this.
5567         * substring-locations.h (class format_string_diagnostic_t): New
5568         class.
5569         (format_warning_va): Convert to
5570         format_string_diagnostic_t::emit_warning_va.
5571         (format_warning_n_va): Convert to
5572         format_string_diagnostic_t::emit_warning_n_va.
5573         (format_warning_at_substring): Convert to
5574         format_string_diagnostic_t::emit_warning.
5575         (format_warning_at_substring_n): Convert to
5576         format_string_diagnostic_t::emit_warning_n.
5578 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
5579             Bernd Schmidt <bernds_cb1@t-online.de>
5581         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
5582         SImode args.
5584 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
5586         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
5587         operand 0 predicate to nonimmediate operand.
5588         (rint<mode>2_frndint): Remove insn pattern.
5589         (rint<mode>2): Change operand 1 predicate to general_operand.
5590         Extend operand 1 to XFmode and generate rintxf2 insn.
5591         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
5592         Do not use X87MODEF mode macro.
5593         (frndintxf2_<rounding>_i387): Rename from
5594         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
5595         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
5596         to XFmode and generate significandxf3 insn.
5598 2018-09-17  Richard Biener  <rguenther@suse.de>
5600         PR tree-optimization/87328
5601         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
5602         visiting unexecutable backedges when not iterating.
5603         (do_rpo_vn): Mark all edges not executable even when not
5604         iterating.
5606 2018-09-17  Martin Jambor  <mjambor@suse.cz>
5608         PR c/63886
5609         * doc/invoke.texi (Warning Options): Likewise.
5611 2018-09-17  Richard Biener  <rguenther@suse.de>
5613         PR tree-optimization/87301
5614         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
5615         clean EH info from leftover copy assignments.
5617 2018-09-17  Martin Liska  <mliska@suse.cz>
5619         PR gcov-profile/85871
5620         * gcov.c (output_intermediate_file): Fix out of bounds
5621         access.
5623 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
5625         * config/arc/arc.c: Object attributes for core4 not reflected
5626         correctly.
5627         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
5628         core3).
5630 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
5632         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
5634 2018-09-17  Martin Liska  <mliska@suse.cz>
5636         * doc/gcov.texi: Document new option --use-hotness-colors.
5637         * gcov.c (struct source_info): Declare new field.
5638         (source_info::source_info): Set default for maximum_count.
5639         (print_usage): Add new -q option.
5640         (process_args): Process it.
5641         (accumulate_line_info): Save src->maximum_count.
5642         (output_line_beginning): Make color line number if
5643         flag_use_hotness_colors is set.
5644         (output_line_details): Pass default argument value.
5645         (output_lines): Pass src->maximum_count.
5647 2018-09-17  Martin Liska  <mliska@suse.cz>
5649         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
5650         Use processor_names table.
5651         * config/i386/i386.c (ix86_default_align): Use
5652         processor_cost_table for alignment values.
5653         (ix86_option_override_internal): Use processor_names.
5654         (ix86_function_specific_print): Likewise.
5655         * config/i386/i386.h (struct processor_costs):
5656         Add alignment values.
5657         (struct ptt): Remove and replace with const char *.
5658         * config/i386/x86-tune-costs.h (struct processor_costs):
5659         Declare default alignments for all costs.
5661 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
5663         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
5664         symbolics or VR_VARYING ranges for ABS_EXPR.
5665         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
5666         when range will wrap.
5668 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
5670         PR middle-end/86864
5671         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
5672         before and after a JUMP_TABLE_DATA.
5674 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
5676         PR middle-end/87188
5677         * dojump.c (do_compare_and_jump): Canonicalize function pointers
5678         when one operand is a function pointer.  Use POINTER_TYPE_P and
5679         FUNC_OR_METHOD_TYPE_P.
5680         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
5681         * fold-const.c (build_range_check): Likewise.
5682         * match.pd (simple_comparison): Likewise.
5684 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
5686         PR c/82967
5687         * spellcheck.c (get_edit_distance_cutoff): New function.
5688         (selftest::test_edit_distance_unit_test_oneway): Rename to...
5689         (selftest::test_get_edit_distance_one_way): ...this.
5690         (selftest::test_get_edit_distance_unit): Rename to...
5691         (selftest::test_get_edit_distance_both_ways): ...this.
5692         (selftest::test_edit_distances): Move tests to this new function,
5693         and test some more pairs of strings.  Update for above renaming.
5694         (selftest::get_old_cutoff): New function.
5695         (selftest::test_get_edit_distance_cutoff): New function.
5696         (selftest::assert_suggested_for): New function.
5697         (ASSERT_SUGGESTED_FOR): New macro.
5698         (selftest::assert_not_suggested_for): New function.
5699         (ASSERT_NOT_SUGGESTED_FOR): New macro.
5700         (selftest::test_suggestions): New function.
5701         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
5702         tests to selftest::test_edit_distances and call it.  Add calls to
5703         selftest::test_get_edit_distance_cutoff and
5704         selftest::test_suggestions.
5705         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
5706         (best_match::consider): Replace hard-coded cutoff calculation with
5707         a call to...
5708         (best_match::get_cutoff): New declaration.
5709         (best_match::get_best_meaningful_candidate): Likewise.
5711 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5713         * builtins.c (fold_builtin_strlen): Remove TODO comment.
5715 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5717         revert:
5718         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5720         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
5721         terminated string literal.
5723 2018-09-14  Martin Sebor  <msebor@redhat.com>
5725         * builtins.c (unterminated_array): Handle ARRAY_REF.
5726         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
5727         * builtins.h (unterminated_array): Declare extern.
5728         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
5729         arrays.
5730         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
5731         calls.
5733 2018-09-14  Martin Sebor  <msebor@redhat.com>
5734             Jeff Law  <law@redhat.com>
5736         * builtins.c (unterminated_array): New.
5737         (expand_builtin_strcpy): Adjust.
5738         (expand_builtin_strcpy_args): Detect unterminated arrays.
5739         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
5740         unterminated arrays.
5741         * gimple-fold.h (get_maxval_strlen): Add argument.
5742         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
5744         * gimple-fold.c (get_range_strlen): Add argument.
5745         (get_maxval_strlen): Adjust.
5746         * gimple-fold.h (get_range_strlen): Add argument.
5748 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
5750         * config/i386/movdirintrin.h: Fix copyright year.
5752 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
5754         * reg-stack.c: Include regs.h.
5755         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
5756         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
5757         FIRST_STACK_REG, not DFmode.
5758         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
5759         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
5760         (change stack): Default register mode to the reg_raw_mode of
5761         FIRST_STACK_REG, not DFmode.
5762         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
5763         (*swapxf): Rename from swapxf.
5765 2018-09-14  Carl Love  <cel@us.ibm.com>
5767         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
5768         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
5770 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
5772         PR target/87224
5773         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
5774         alternatives.
5776 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
5778         PR target/85628
5779         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
5781 2018-09-14  Jason Merrill  <jason@redhat.com>
5783         Fix --enable-gather-detailed-mem-stats.
5784         * hash-table.c (hash_table_usage): Change from variable to function.
5785         * hash-table.h: Adjust.
5786         * Makefile.in: Add missing dependencies on hash-table.h.
5788 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5790         PR tree-optimization/87259
5791         PR lto/87283
5792         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
5793         execute_cse_reciprocals_1 has tried transforming.
5795 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
5797         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
5798         VR_VARYING for PLUS/MINUS_EXPR.
5800 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
5802         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
5803         formatting.
5805 2018-09-14  Richard Biener  <rguenther@suse.de>
5807         PR middle-end/63155
5808         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
5809         bits for the merged partition.
5811 2018-09-13  Martin Sebor  <msebor@redhat.com>
5812             Bernd Edlinger  <bernd.edlinger@hotmail.de>
5814         * builtins.h (c_srlen): Add argument.
5815         * builtins.c (warn_string_no_nul): New function.
5816         (c_strlen): Add argument and use it.  Update recursive calls.
5817         Pass DECL argument to string_constant to get info on non
5818         terminated strings.  Update *NONSTR as needed.
5819         (fold_builtin_strlen): Add argument to calls to c_strlen.
5820         Warn for unterminated arrays.
5821         (warn_string_no_null): Add prototype.
5822         * expr.c (string_constant): Update arguments.  Update recursive
5823         calls appropriately.  Detect missing NUL terminator and outermost
5824         declaration its missing in.
5825         Improve checks for arrays with nonzero lower bound or elements
5826         that are not a single byte.  Simplify offset computation.
5827         Simplify checks for non-NUL terminated strings.
5828         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
5829         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
5831 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5833         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
5834         correctly.
5835         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
5836         argument.  Fix range checks.
5837         * fold-const.h (c_getstr): Adjust protoype.
5838         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
5839         string is constant but contains no NUL byte.
5841         * expr.c (string_constant): Adjust function comment.
5842         Remove bogus check for zero termination.
5844         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
5846         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
5847         (get_constant_size): Don't make STRING_CSTs larger than they are.
5848         (check_string_literal): New check function for STRING_CSTs.
5849         (output_constant): Use it.
5851 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
5853         PR target/86812
5854         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
5856 2018-09-13  Richard Biener  <rguenther@suse.de>
5858         PR tree-optimization/87263
5859         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
5860         (struct unwind_state): Add max_rpo field.
5861         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
5862         Compute max_rpo, the max RPO number a block can be backwards reached
5863         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
5864         separating it from the iterating mode.
5866 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
5868         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
5869         (rfs_decision): New scheduling decision.
5871 2018-09-13  Richard Biener  <rguenther@suse.de>
5873         PR bootstrap/87134
5874         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
5875         (vn_nary_op_insert_pieces_predicated): Do not write useless
5876         valid_dominated_by_p entry outside of the allocated storage.
5878 2018-09-13  Omar Sandoval  <osandov@osandov.com>
5879             Tom de Vries  <tdevries@suse.de>
5881         PR debug/86985
5882         * dwarf2out.c (is_c): New function.
5883         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
5885 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
5887         PR target/85628
5888         * config/aarch64/aarch64.md (*aarch64_bfxil):
5889         Define.
5890         * config/aarch64/constraints.md (Ulc): Define.
5891         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
5892         Define.
5893         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
5894         New function.
5896 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
5898         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
5899         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
5900         aarch64_layout_frame call.
5901         (aarch64_expand_epilogue): Likewise.
5902         (aarch64_initial_elimination_offset): Likewise.
5903         (aarch64_get_separate_components): Likewise.
5904         (aarch64_use_return_insn_p): Likewise.
5905         (aarch64_layout_frame): Remove unneeded check.
5907 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
5909         * configure.ac: Only append
5910         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
5911         gcc_config_arguments if it was never reconfigured or last reconfigure
5912         was with different arguments.
5913         * configure: Regenerated.
5915 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
5916             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5918         PR middle-end/87290
5919         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
5920         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
5922 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
5924         PR tree-optimization/87287
5925         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
5926         X % C == 0 to X % (unsigned) C == 0 optimization to ...
5927         * match.pd (X % C == 0): ... here.  New optimization.
5929 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
5931         PR middle-end/82853
5932         * expr.h (maybe_optimize_mod_cmp): Declare.
5933         * expr.c (mod_inv): New function.
5934         (maybe_optimize_mod_cmp): New function.
5935         (do_store_flag): Use it.
5936         * cfgexpand.c (expand_gimple_cond): Likewise.
5938 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
5939             Julian Brown  <julian@codesourcery.com>
5941         PR middle-end/86336
5942         * gimplify.c (gimplify_scan_omp_clauses): Set
5943         target_firstprivatize_array_bases in OpenACC parallel and kernels
5944         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
5945         OpenACC data regions.
5947 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
5949         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
5950         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
5952 2018-09-12  Richard Biener  <rguenther@suse.de>
5954         PR tree-optimization/87280
5955         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
5956         edge but unreachable target.
5957         (do_rpo_vn): For conservatively handling a PHI only mark
5958         the backedge executable but not the block reachable.
5960 2018-09-12  Richard Biener  <rguenther@suse.de>
5962         PR tree-optimization/87266
5963         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
5964         visited blocks.
5966 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
5968         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
5969         constants.
5970         ("trunc<BFP:mode><DFP_ALL:mode>2")
5971         ("trunc<DFP_ALL:mode><BFP:mode>2")
5972         ("extend<BFP:mode><DFP_ALL:mode>2")
5973         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
5974         according to the target operand type.
5976 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
5977             Andreas Krebbel  <krebbel@linux.ibm.com>
5979         PR tree-optimization/86844
5980         * gimple-ssa-store-merging.c
5981         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
5982         there are any overlapping stores in between them, make sure they are
5983         also coalesced or we give up completely.
5985 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
5987         PR middle-end/87248
5988         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
5989         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
5991 2018-09-12  Tom de Vries  <tdevries@suse.de>
5993         * common.opt (gdescribe-dies): Add option.
5994         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
5995         attribute for artifical and nameless decls.
5996         (dwarf2out_register_external_die): Add description attribute to
5997         external reference die.
5998         (add_desc_attribute): New functions.
5999         (gen_subprogram_die): Add description attribute to
6000         DW_TAG_call_site_parameter.
6001         * tree-pretty-print.c (print_generic_expr_to_str): New function.
6002         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
6003         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
6004         -gno-describe-dies.
6005         (@item -gdescribe-dies): Add.
6007 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
6009         * tree-vrp.c (vrp_shift_undefined_p): Remove.
6010         (extract_range_from_binary_expr_1: Call
6011         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
6012         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
6013         depend on sign.
6015 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
6017         * gimple-ssa-warn-alloca.c
6018         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
6019         field for ALLOCA_BOUND_*_LARGE.
6021 2018-09-11  Nathan Sidwell  <nathan@acm.org>
6023         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
6025 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
6027         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
6028         for clobbers.  Remove obsolete comment.
6030 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
6032         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
6033         mpxchk, mpxld and mpxst types.
6034         (define_attr length_immediate): Remove all processing of mpx types.
6035         (define_attr prefix_0f): Ditto.
6036         (define_attr memory): Ditto.
6038 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
6040         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
6041         (log<mode>2): Change operand 1 predicate to general_operand.
6042         Extend operand 1 to XFmode and generate logxf3 insn.
6043         (log10<mode>2): Change operand 1 predicate to general_operand.
6044         Extend operand 1 to XFmode and generate log10xf3 insn.
6045         (log2<mode>2): Change operand 1 predicate to general_operand.
6046         Extend operand 1 to XFmode and generate log2xf3 insn.
6047         (fyl2xp1_extend<mode>xf3_i387): Remove.
6048         (log1p<mode>2): Change operand 1 predicate to general_operand.
6049         Extend operand 1 to XFmode and generate log1pxf3 insn.
6050         (fxtract_extend<mode>xf3_i387): Remove.
6051         (logb<mode>2): Change operand 1 predicate to general_operand.
6052         Extend operand 1 to XFmode and generate logbxf3 insn.
6053         (ilogb<mode>2): Change operand 1 predicate to general_operand.
6054         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
6055         (significand<mode>2): Change operand 1 predicate to general_operand.
6056         Extend operand 1 to XFmode and generate significandxf3 insn.
6058 2018-09-11  Nathan Sidwell  <nathan@acm.org>
6060         * gcc.c (perror_with_name, pfatal_with_name): Delete.
6061         (load_specs): Use fatal_error.
6062         (DELETE_IF_ORDINARY, process_command): Use error.
6063         (execute, run_attempt): Use fatal_error.
6065 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
6067         * diagnostic-core.h (sorry_at): New prototype.
6068         * diagnostic.c (sorry_at): New function.
6070 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
6072         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
6073         by zero as VR_UNDEFINED.
6075 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
6077         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
6078         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
6079         (<sincos>mode2): New expander.
6080         (sincos_extend<mode>xf3_i387): Remove insn pattern.
6081         (sincos -> sin, cos splitters): Remove splitter patterns.
6082         (sincos<mode>3): Change operand 2 predicate to general_operand.
6083         Extend operand 2 to XFmode and generate sincosxf3 insn.
6084         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
6085         Change operand 3 predicate to const1_operand.
6086         (fptan_extend<mode>xf4_i387): Remove insn pattern.
6087         (tanxf2): Update operands in the call to fptanxf4_i387.
6088         (tan<mode>2): Change operand 1 predicate to general_operand.
6089         Extend operand 1 to XFmode and generate tanxf3 insn.
6090         (atan2xf3): Rename from *fpatanxf3_i387.
6091         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
6092         (atan2xf3): Remove expander.
6093         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
6094         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
6095         (atan<mode>2): Change operand 1 predicate to general_operand.
6096         Extend operand 1 to XFmode and generate atanxf3 insn.
6098 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
6100         * config/i386/i386.md (x87/SSE constant load splitter): Use
6101         memory_operand instead of nonimmediate_operand for input operand
6102         predicate.
6104 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
6106         * config/i386/i386.md (float partial SSE register stall splitter): Move
6107         splitter near its instruction pattern.
6108         (float_extend partial SSE register stall splitter): Ditto.
6109         (float_truncate partial SSE register stall splitter): Ditto.
6111 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
6113         PR target/86794
6114         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6115         to speculation_safe_value_not_needed.
6117         PR target/85666
6118         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
6119         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
6120         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
6121         leaf_function_p, instead use has_hard_reg_initial_val.
6123 2018-09-09  Nathan Sidwell  <nathan@acm.org>
6125         * gcc.h (pfatal_with_name): Don't declare here.
6126         * gcc.c (pfatal_with_name): Make static.
6128 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
6130         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
6131         earlyclobber.
6133 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
6135         PR rtl-optimization/85458
6136         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
6137         priority hook to reduce the priority of EXPR.
6139 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
6141         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
6142         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
6143         with X87_ENABLE_FLOAT.  Remove preparation code.
6144         (*float<SWI48:mode><MODEF:mode>2): Rename from
6145         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
6146         math using "enabled" attribute.
6147         (*floatdi<MODEF:mode>2_i387): Rename from
6148         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
6149         enable for 32bit targets only.
6150         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
6151         splitter.
6152         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
6153         as operand 1 predicate.  Rewrite as define_insn_and_split.
6154         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
6156 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
6158         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
6159         to fallthru to FLOAT case.
6161 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
6163         PR target/86731
6164         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
6165         around folding of vec_sl to handle out of range shift values.
6167 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
6169         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
6170         Update callers to gen_fix_trunc<mode>_i387_fisttp
6171         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
6172         nonimmediate_operand.
6173         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
6174         and corresponding splitters.
6175         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
6176         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
6177         (fix_truncdi_i387_with_temp): Remove insn pattern
6178         and corresponding splitters.
6179         (fix_trunc<mode>_i387): Change operand 0 predicate to
6180         nonimmediate_operand.
6181         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
6182         and corresponding splitters.
6183         (*fistdi2_1): Remove.
6184         (fistdi2): Ditto.
6185         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
6186         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
6187         (*fist<mode>2_1): Remove.
6188         (fist<mode>2): Ditto.
6189         (fist<mode>2_with_temp): Remove insn pattern and corresponding
6190         splitters.
6191         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
6192         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
6193         (fistdi2_<rounding>): Change operand 0 predicate to
6194         nonimmediate_operand.
6195         (fistdi2_<rounding>_with_temp): Remove insn pattern
6196         and corresponding splitters.
6197         (fist<mode>2_<rounding>): Change operand 0 predicate to
6198         nonimmediate_operand.
6199         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
6200         and corresponding splitters.
6202         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
6204 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6206         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
6207         the init value.
6209 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
6211         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
6212         early gimple folding of vec_splat().
6213         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
6214         * gimple-fold.h: Add an extern define for tree_vec_extract().
6216 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
6218         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
6219         wrappers around TREE_TYPE comparisons.
6221 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
6223         PR target/80080
6224         * config/s390/predicates.md: Add nonsym_memory_operand.
6225         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
6226         contains a SYMBOL_REF, load it into an intermediate pseudo.
6227         (s390_emit_compare_and_swap): Legitimize operand.
6228         * config/s390/s390.md: Use the new nonsym_memory_operand
6229         with UNSPECV_CAS patterns.
6231 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
6233         PR target/80080
6234         * config/s390/s390-passes.def: New file.
6235         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
6236         declaration.
6237         (make_pass_s390_early_mach): Add declaration.
6238         * config/s390/s390.c (make_pass_s390_early_mach):
6239         (s390_option_override): Remove dynamic registration.
6240         * config/s390/t-s390: Add s390-passes.def.
6242 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
6244         * config/s390/s390.c (s390_decompose_constant_pool_ref):
6245         Remove UNSPEC_LTREL_BASE check.
6246         (annotate_constant_pool_refs): Likewise.
6247         (find_constant_pool_ref): Likewise.
6248         (find_ltrel_base): Removed.
6249         (replace_ltrel_base): Removed.
6250         (s390_mainpool_finish): Remove replace_ltrel_base call.
6251         (s390_chunkify_start): Remove pending LTREL_BASE logic.
6252         (s390_chunkify_finish): Remove replace_ltrel_base call.
6253         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
6255 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
6257         PR target/86779
6258         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6259         to speculation_safe_value_not_needed.
6261 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
6262             Bernd Schmidt  <bernds_cb1@t-online.de>
6264         * config/nvptx/nvptx-opts.h: New file.
6265         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
6266         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
6267         (ASM_SPEC): Define.
6268         (TARGET_SM35): New macro.
6269         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
6270         correct predicate.
6271         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
6272         values.
6273         (misa=): New option.
6274         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
6276 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
6278         * config/i386/i386.md (truncdfsf2): Remove expander.
6279         (truncdfsf2_with_temp): Ditto.
6280         (truncxf<mode>2): Ditto.
6281         (*truncdfsf_fast_mixed): Remove insn pattern.
6282         (*truncdfsf_fast_i387): Ditto.
6283         (*truncdfsf_mixed): Ditto.
6284         (*truncdfsf_i387): Ditto.
6285         (*truncdfsf2_i387_1): Ditto.
6286         (*truncxfsf2_mixed): Ditto.
6287         (*truncxfdf2_mixed): Ditto.
6288         (*truncxf<mode>2_i387_noop): Ditto. Update callers
6289         to call gen_truncxf<mode>2 instead.
6290         (*truncxf<mode>2_i387): Remove.
6291         (reg->reg splitters): Remove splitter pattern.
6292         (reg->mem splitters): Ditto.
6294         (truncdfsf2): New insn pattern.
6295         (truncxf<mode>2): Ditto.
6297 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6299         * tree-ssa-math-opts.c (is_mult_by): New function.
6300         (is_square_of): Use the above.
6301         (optimize_recip_sqrt): New function.
6302         (pass_cse_reciprocals::execute): Use the above.
6304 2018-09-05  Richard Biener  <rguenther@suse.de>
6306         PR bootstrap/87134
6307         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
6308         to zero-init the emplaced vec.
6310 2018-09-05  Martin Liska  <mliska@suse.cz>
6312         PR tree-optimization/87205
6313         * tree-switch-conversion.c (pass_lower_switch::execute):
6314         Group cases for switch statements.
6316 2018-09-05  Richard Biener  <rguenther@suse.de>
6318         PR tree-optimization/87217
6319         * tree-ssa-sccvn.c (vuse_valueize): New.
6320         (vn_reference_lookup_pieces): Use it.
6321         (vn_reference_lookup): Likewise.
6323 2018-09-05  Nathan Sidwell  <nathan@acm.org>
6325         PR c++/87137
6326         * stor-layout.c (place_field): Scan forwards to check last
6327         bitfield when ms_bitfield_placement is in effect.
6329 2018-09-05  Richard Biener  <rguenther@suse.de>
6331         PR bootstrap/87225
6332         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
6333         return.
6335 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
6336             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
6338         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
6339         * config.gcc (extra_objs): Build it.
6340         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
6341         Likewise.
6342         * config/aarch64/aarch64-passes.def
6343         (pass_tag_collision_avoidance): New pass.
6344         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
6345         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
6346         (aarch64_classify_address): Remove static qualifier.
6347         (aarch64_address_info, aarch64_address_type): Move to...
6348         * config/aarch64/aarch64-protos.h: ... here.
6349         (make_pass_tag_collision_avoidance): New function.
6350         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
6351         New tuning flag.
6353 2018-09-05  Martin Liska  <mliska@suse.cz>
6355         * doc/gcov.texi: Update documentation of humar
6356         readable mode.
6357         * gcov.c (format_count): Print one decimal place, it provides
6358         more fine number of situations like '1G' vs. '1.4G'.
6360 2018-09-05  Martin Liska  <mliska@suse.cz>
6362         PR target/87164
6363         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
6364         * optc-gen.awk: Allow 'Var' for Deprecated options in order
6365         to generate a MASK value.
6367 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
6369         PR debug/86593
6370         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
6371         if frame pointer isn't used.
6372         (compute_frame_pointer_to_fb_displacement): Likewise.
6374 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
6376         PR target/87198
6377         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
6378         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
6379         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
6380         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
6381         and OPTION_MASK_ISA_XSAVEC_UNSET.
6383 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
6385         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
6386         XOR operations in NAND case.
6388 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
6390         * wide-int-range.cc (wide_int_range_convert): New.
6391         * wide-int-range.h (wide_int_range_convert): New.
6392         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
6393         code into wide_int_range_convert.
6394         (extract_range_into_wide_ints): Do not munge anti range constants
6395         into the entire domain.  Just return the range back.
6397 2018-09-04  Martin Liska  <mliska@suse.cz>
6399         * genmatch.c (output_line_directive): Add new argument
6400         fnargs.
6401         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
6403 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
6405         * doc/invoke.texi (Option Summary): Add whitespace.
6407         * doc/invoke.texi (Option Summary): Add -Waligned-new.
6409 2018-09-04  Richard Biener  <rguenther@suse.de>
6411         PR tree-optimization/87211
6412         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
6413         backedge value we're supposed to treat as VARYING also number
6414         the PHI to VARYING in case it got a different value-number already.
6416 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
6418         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
6419         (extract_range_from_binary_expr_1): Do not call
6420         vrp_can_optimize_bit_op.
6421         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
6422         static.
6423         (wide_int_range_get_mask_and_bounds): New.
6424         (wide_int_range_optimize_bit_op): New.
6425         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
6426         (wide_int_range_bit_and): Same.
6427         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
6428         (wide_int_range_optimize_bit_op): New.
6429         (wide_int_range_get_mask_and_bounds): New.
6431 2018-09-04  Richard Biener  <rguenther@suse.de>
6433         PR tree-optimization/87176
6434         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
6435         variable.  When value-numbering a virtual PHI node make sure
6436         to not value-number to the backedge value.
6438 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
6440         * doc/extend.texi (Long Long, Hex Floats): Document support for
6441         long long and hex floats in more recent versions of ISO C++.
6443 2018-09-03  Richard Biener  <rguenther@suse.de>
6445         PR tree-optimization/87177
6446         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
6447         cleanup.
6449 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
6451         * bb-reorder.c (edge_order): Convert to C-qsort-style
6452         tri-state comparator.
6453         (reorder_basic_blocks_simple): Change std::stable_sort to
6454         gcc_stablesort.
6456 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
6458         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
6459         tri-state comparator.
6460         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
6462 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
6464         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
6465         (mergesort): ... here as maximum count for using netsort.
6466         (gcc_qsort): Set nlim to 3 if stable sort is requested.
6467         (gcc_stablesort): New.
6468         * system.h (gcc_stablesort): Declare.
6470 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
6472         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
6473         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
6474         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
6476 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
6478         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
6479         lxsdx and stxsdx alternatives.
6480         (*mov<mode>_hardfloat64): Ditto.
6481         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
6483 2018-09-03  Richard Biener  <rguenther@suse.de>
6485         PR tree-optimization/87200
6486         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
6487         simplify result.
6489 2018-09-03  Martin Liska  <mliska@suse.cz>
6491         PR tree-optimization/87201
6492         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
6493         Fix parenthesis in an expression.
6495 2018-09-03  Richard Biener  <rguenther@suse.de>
6497         PR tree-optimization/87197
6498         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
6499         visited.  CSE the VN_INFO hashtable lookup.
6501         PR tree-optimization/87169
6502         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
6503         iterating make sure there's no extra backedges from irreducible
6504         regions feeding the header.  Mark the destination block
6505         executable.
6507 2018-09-03  Martin Liska  <mliska@suse.cz>
6509         PR driver/83193
6510         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
6511         * common/common-targhooks.c (default_get_valid_option_values):
6512         New function.
6513         * common/common-targhooks.h (default_get_valid_option_values):
6514         Likewise.
6515         * common/config/i386/i386-common.c: Move processor_target_table
6516         from i386.c.
6517         (ix86_get_valid_option_values): New function.
6518         (TARGET_GET_VALID_OPTION_VALUES): New macro.
6519         * config/i386/i386.c (struct ptt): Move to i386-common.c.
6520         (PTA_*): Move all defined masks into i386-common.c.
6521         (ix86_function_specific_restore): Use new processor_cost_table.
6522         * config/i386/i386.h (struct ptt): Moved from i386.c.
6523         (struct pta): Likewise.
6524         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
6525         * doc/tm.texi.in: Likewise.
6526         * opt-suggestions.c (option_proposer::suggest_option):
6527         Pass prefix to build_option_suggestions.
6528         (option_proposer::get_completions): Likewise.
6529         (option_proposer::build_option_suggestions): Use the new target
6530         hook.
6531         * opts.c (struct option_help_tuple): New struct.
6532         (print_filtered_help): Use the new target hook.
6534 2018-09-03  Martin Liska  <mliska@suse.cz>
6536         PR middle-end/59521
6537         * predict.c (set_even_probabilities): Add likely_edges
6538         argument and handle cases where we have precisely one
6539         likely edge.
6540         (combine_predictions_for_bb): Catch also likely_edges.
6541         (tree_predict_by_opcode): Handle gswitch statements.
6542         * tree-cfg.h (find_case_label_for_value): New declaration.
6543         (find_taken_edge_switch_expr): Likewise.
6544         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
6545         Find pivot in decision tree based on probabily, not by number of
6546         nodes.
6548 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
6550         * doc/standards.texi (Standards): Update Objective-C reference.
6552 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
6554         * doc/install.texi (Prerequisites): Update link for MPC.
6556 2018-09-01  Michael Matz  <matz@suse.de>
6558         PR tree-optimization/87074
6559         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
6560         PHIs for outer-loop uses.
6562 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
6564         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
6565         * doc/invoke.texi (C Dialect Options): Ditto.
6567 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
6569         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
6571 2018-08-31  Richard Biener  <rguenther@suse.de>
6573         PR tree-optimization/87168
6574         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
6575         (rpo_elim::eliminate_avail): When OP was not visited it must
6576         be available.
6578 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
6580         * tree-vrp.c (copy_value_range): Convert param "from" from
6581         "value_range *" to "const value_range *".
6582         (range_is_null): Likewise for param "vr".
6583         (range_int_cst_p): Likewise.
6584         (range_int_cst_singleton_p): Likewise.
6585         (symbolic_range_p): Likewise.
6586         (value_ranges_intersect_p): Likewise for both params.
6587         (value_range_nonnegative_p): Likewise for param "vr".
6588         (value_range_constant_singleton): Likewise.
6589         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
6590         (extract_range_into_wide_ints): Likewise for param "vr".
6591         (extract_range_from_multiplicative_op): Likewise for params "vr0"
6592         and "vr1".
6593         (vrp_can_optimize_bit_op): Likewise.
6594         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
6595         "vr1_".
6596         (extract_range_from_unary_expr): Likewise.
6597         (debug_value_range): Likewise for param "vr".
6598         (value_range::dump): Add "const" qualifier.
6599         (vrp_prop::check_array_ref): Convert local "vr" from
6600         "value_range *" to "const value_range *".
6601         (vrp_prop::check_mem_ref): Likewise.
6602         (vrp_prop::visit_stmt): Likewise for local "old_vr".
6603         (vrp_intersect_ranges_1): Likewise for param "vr_1".
6604         (vrp_intersect_ranges): Likewise.
6605         (simplify_stmt_for_jump_threading): Likewise for local "vr".
6606         (vrp_prop::vrp_finalize): Likewise.
6607         * tree-vrp.h (value_range::dump): Add "const" qualifier.
6608         (vrp_intersect_ranges): Add "const" qualifier to params as above.
6609         (extract_range_from_unary_expr): Likewise.
6610         (value_range_constant_singleton): Likewise.
6611         (symbolic_range_p): Likewise.
6612         (copy_value_range): Likewise.
6613         (extract_range_from_binary_expr_1): Likewise.
6614         (range_int_cst_p): Likewise.
6615         (vrp_set_zero_nonzero_bits): Likewise.
6616         (range_int_cst_singleton_p): Likewise.
6618 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
6620         * config/aarch64/arm_neon.h (vabsd_s64): New.
6621         (vnegd_s64): Likewise.
6623 2018-08-31  Martin Jambor  <mjambor@suse.cz>
6625         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
6627 2018-08-31  Martin Liska  <mliska@suse.cz>
6629         * ipa-icf.c (sem_item::add_type): Use
6630         sem_item::m_type_hash_cache.
6631         * ipa-icf.h: Move the cache from sem_item_optimizer
6632         to sem_item.
6634 2018-08-31  Nathan Sidwell  <nathan@acm.org>
6636         * doc/extend.texi (Backwards Compatibility): Remove implicit
6637         extern C leeway of () being (...).
6639 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6641         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
6643 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
6645         PR target/86684
6646         PR target/87149
6647         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
6649 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
6651         PR middle-end/87138
6652         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
6653         gen_int_mode.  Formatting fixes.
6655 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
6657         * target.def (custom_function_descriptors): Improve documentation.
6658         * doc/tm.texi.in (Trampolines): Expand discussion of function
6659         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
6660         beginning of the section.
6661         * doc/tm.texi: Regenerated.
6663 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
6665         * cfg.h (class auto_edge_flag): Spell out the template-id of the
6666         base class in the initializer list.  This is a workaround for
6667         building with older GCC.
6668         (class auto_bb_flag): Likewise.
6670 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
6672         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
6673         (altivec_vcmpequ<VI_char>_p): Remove star.
6674         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
6675         vector load modes.
6676         (expand_strncmp_vec_sequence): New function.
6677         (emit_final_str_compare_vec): New function.
6678         (expand_strn_compare): Add support for vector strncmp.
6679         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
6680         length specification to bytes.
6681         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
6682         (vcmpnezb_p): New pattern.
6683         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
6684         for option -mstring-compare-inline-limit.
6686 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
6688         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
6689         (PTA_SKYLAKE): Add PTA_AES.
6690         (PTA_GOLDMONT): Likewise.
6692 2018-08-29  Jan Hubicka  <jh@suse.cz>
6694         PR lto/86517
6695         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
6696         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
6698 2018-08-29  Jan Hubicka  <jh@suse.cz>
6700         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
6701         TYPE_STUB_DECL.
6702         (hash_tree): Do not visit TYPE_STUB_DECL.
6703         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
6704         stream TYPE_STUB_DECL.
6705         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
6706         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
6707         after free_lang_data.
6708         (type_in_anonymous_namespace_p): Likewise.
6710 2018-08-29  Jan Hubicka  <jh@suse.cz>
6712         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
6713         comment that it has to be even number.
6714         (class sreal): Change m_sig type to int32_t.
6715         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
6716         int64_t for temporary calculations.
6717         (sreal_verify_basics): Drop one bit from minimum and maximum.
6719 2018-08-30  Richard Biener  <rguenther@suse.de>
6721         PR tree-optimization/87147
6722         * tree-ssa-sccvn.c (SSA_VISITED): New function.
6723         (visit_phi): When the degenerate result is from the backedge and
6724         we didn't visit its definition yet drop to VARYING.
6725         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
6727 2018-08-29  Jan Hubicka  <jh@suse.cz>
6729         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
6730         DECL_VINDEX.
6731         (hash_tree): Likewise.
6733 2018-08-29  Jan Hubicka  <jh@suse.cz>
6735         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
6736         and TYPE_NEXT_REF_TO.
6738 2018-08-29  Jan Hubicka  <jh@suse.cz>
6740         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
6741         comment that it has to be even number.
6742         (class sreal): Change m_sig type to int32_t.
6743         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
6744         int64_t for temporary calculations.
6745         (sreal_verify_basics): Drop one bit from minimum and maximum.
6747 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
6749         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
6751 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
6753         PR middle-end/86995
6754         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
6755         if to_add is negative.
6757 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6759         PR middle-end/87053
6760         * builtins.c (c_strlen): Improve range checks.
6762 2018-08-29  Martin Sebor  <msebor@redhat.com>
6763             Jeff Law  <law@redhat.com>
6765         PR tree-optimization/86714
6766         PR tree-optimization/86711
6767         * builtins.c (c_strlen): Add arguments to call to string_constant.
6768         * expr.c (string_constant): Add argument.  Detect missing nul
6769         terminator and outermost declaration it's missing in.
6770         * expr.h (string_constant): Add argument.
6771         * fold-const.c (read_from_constant_string): Add arguments to call to
6772         string_constant.
6773         (c_getstr): Likewise.
6774         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
6775         to string_constant.
6776         * tree-ssa-strlen.c (get_stridx): Likewise.
6778 2018-08-29  Jan Hubicka  <jh@suse.cz>
6780         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
6781         Do not stream DECL_VINDEX.
6782         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
6783         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
6784         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
6786 2018-08-29  Richard Biener  <rguenther@suse.de>
6788         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
6789         virtual operands that are not default defs to honor region
6790         boundaries.
6791         (rpo_vn_valueize): Remove ineffective code here.
6793 2018-08-29  Richard Biener  <rguenther@suse.de>
6795         PR tree-optimization/87132
6796         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
6797         when skipping defs reachable over backedges.
6799 2018-08-29  Richard Biener  <rguenther@suse.de>
6801         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
6802         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
6803         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
6804         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
6805         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
6806         (vn_reference_lookup_3): Remove use of const_parms.
6807         (free_rpo_vn): Do not free const_parms.
6808         (do_rpo_vn): Do not call init_const_parms.
6809         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
6810         SSA_NAME_POINTS_TO_READONLY_MEMORY.
6811         (call_may_clobber_ref_p_1): Likewise.
6813 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
6815         PR other/86726
6816         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
6817         (-O): Ditto.
6818         (-ftree-scev-cprop): Document.
6820 2018-08-29  Jan Hubicka  <jh@suse.cz>
6822         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
6823         parameters.
6824         (sreal constructor): Update.
6825         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
6826         sreal:operator/): Update.
6828 2018-08-29  Martin Liska  <mliska@suse.cz>
6830         * tree-switch-conversion.c (switch_conversion::expand):
6831         Strenghten assumption about gswitch statements.
6833 2018-08-29  Richard Biener  <rguenther@suse.de>
6835         PR tree-optimization/87117
6836         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
6837         re-value-number released SSA VDEFs.
6839 2018-08-29  Richard Biener  <rguenther@suse.de>
6841         PR tree-optimization/87126
6842         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
6844 2018-08-28  Jim Wilson  <jimw@sifive.com>
6846         * config/riscv/pic.md: Rewrite.
6847         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
6848         invalid address.
6849         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
6850         (SOFTF, default_load, softload, softstore): New.
6852 2018-08-28  Jeff Law  <law@redhat.com>
6854         * fold-const.c (fold_binary_loc): Remove recently added assert.
6856 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
6858         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
6859         to OP parmeter of generated function.
6861 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
6863         PR tree-optimization/87009
6864         * match.pd: Add boolean optimizations.
6866 2018-08-28  Martin Sebor  <msebor@redhat.com>
6868         PR middle-end/86631
6869         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
6870         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
6871         (pass_walloca::gate): Use it.
6872         (alloca_call_type): Same.
6873         (pass_walloca::execute): Same.
6874         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
6876 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
6878         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
6879         GCC_VERSION for usage of "__gcc_dump_printf__" format from
6880         >= 3005 to >= 9000.
6882 2018-08-28  Richard Biener  <rguenther@suse.de>
6884         PR tree-optimization/87124
6885         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
6886         constants before looking up avail.
6888 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
6890         PR middle-end/87099
6891         * calls.c (maybe_warn_nonstring_arg): Punt early if
6892         warn_stringop_overflow is zero.  Don't call get_range_strlen
6893         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
6894         Swap comparison operands to have constants on rhs.  Only use
6895         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
6896         increment lenrng[0].
6898 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
6900         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
6901         use of tree_to_shwi.  Remove duplicated test for the size being
6902         a whole number of bytes.
6904 2018-08-28  Richard Biener  <rguenther@suse.de>
6906         PR tree-optimization/87117
6907         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
6908         Handle removed stmt without LHS (GIMPLE_NOP).
6910 2018-08-28  Richard Biener  <rguenther@suse.de>
6912         PR tree-optimization/87117
6913         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
6914         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
6916 2018-08-28  Richard Biener  <rguenther@suse.de>
6918         PR tree-optimization/87117
6919         * tree-ssa-pre.c (compute_avail): Do not make expressions
6920         with predicated values available.
6921         (get_expr_value_id): Assert we do not run into predicated value
6922         expressions.
6924 2018-08-28  Richard Biener  <rguenther@suse.de>
6926         PR tree-optimization/87117
6927         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
6928         get virtual operands.
6929         (get_expr_operands): Handle STRING_CST like other decls.
6931 2018-08-28  Martin Liska  <mliska@suse.cz>
6933         * tree.h: Update documentation of fndecl_built_in_p
6934         functions.
6937 2018-08-27  Jeff Law  <law@redhat.com>
6938         PR tree-optimization/87110
6939         * tree-ssa-dse.c (compute_trims): Handle non-constant
6940         TYPE_SIZE_UNIT.
6942 2018-08-27  Martin Sebor  <msebor@redhat.com>
6944         PR tree-optimization/86914
6945         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
6947 2018-08-27  Martin Sebor  <msebor@redhat.com>
6949         PR tree-optimization/87112
6950         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
6951         the type of the bound argument.
6953 2018-08-27  Jeff Law  <law@redhat.com>
6955         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
6956         type does not have a TYPE_SIZE_UNIT.
6958 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
6960         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
6961         with include of backend.h.
6963 2018-08-27  Richard Biener  <rguenther@suse.de>
6965         PR tree-optimization/86927
6966         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
6967         use const cond reduction code.
6969 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
6971         PR tree-optimization/85758
6972         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
6974 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
6976         PR c++/87091
6977         * diagnostic-show-locus.c (class layout_range): Update for
6978         conversion of show_caret_p to a tri-state.
6979         (layout_range::layout_range): Likewise.
6980         (make_range): Likewise.
6981         (layout::maybe_add_location_range): Likewise.
6982         (layout::should_print_annotation_line_p): Don't show annotation
6983         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
6984         (layout::get_state_at_point): Update for conversion of
6985         show_caret_p to a tri-state.  Bail out early for
6986         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
6987         underlining or source colorization.
6988         (gcc_rich_location::add_location_if_nearby): Update for conversion
6989         of show_caret_p to a tri-state.
6990         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
6991         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
6992         Likewise.
6993         (selftest::test_one_liner_labels): Likewise.
6994         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
6995         conversion of show_caret_p to a tri-state.
6996         * pretty-print.c (text_info::set_location): Likewise.
6997         * pretty-print.h (text_info::set_location): Likewise.
6998         * substring-locations.c (format_warning_n_va): Likewise.
6999         * tree-diagnostic.c (default_tree_printer): Likewise.
7000         * tree-pretty-print.c (newline_and_indent): Likewise.
7002 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
7004         PR c++/87091
7005         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
7006         line above for line-insertion fix-it hints.
7007         (selftest::test_fixit_insert_containing_newline): Update the
7008         expected results, and add a test with line-numbering enabled.
7010 2018-08-27  Martin Liska  <mliska@suse.cz>
7012         PR sanitizer/86962
7013         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
7014         params with DECL_HAS_VALUE_EXPR_P.
7016 2018-08-27  Martin Liska  <mliska@suse.cz>
7018         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
7019         selected expansion strategy.
7021 2018-08-27  Martin Liska  <mliska@suse.cz>
7023         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
7024         * builtins.c (is_builtin_fn): Likewise.
7025         * attribs.c (diag_attr_exclusions): Use new function
7026         fndecl_built_in_p and remove check for FUNCTION_DECL if
7027         possible.
7028         (builtin_mathfn_code): Likewise.
7029         (fold_builtin_expect): Likewise.
7030         (fold_call_expr): Likewise.
7031         (fold_builtin_call_array): Likewise.
7032         (fold_call_stmt): Likewise.
7033         (set_builtin_user_assembler_name): Likewise.
7034         (is_simple_builtin): Likewise.
7035         * calls.c (gimple_alloca_call_p): Likewise.
7036         (maybe_warn_nonstring_arg): Likewise.
7037         * cfgexpand.c (expand_call_stmt): Likewise.
7038         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
7039         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
7040         (cgraph_node::verify_node): Likewise.
7041         * cgraphclones.c (build_function_decl_skip_args): Likewise.
7042         (cgraph_node::create_clone): Likewise.
7043         * config/arm/arm.c (arm_insert_attributes): Likewise.
7044         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
7045         * dse.c (scan_insn): Likewise.
7046         * expr.c (expand_expr_real_1): Likewise.
7047         * fold-const.c (operand_equal_p): Likewise.
7048         (fold_binary_loc): Likewise.
7049         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
7050         * gimple-low.c (lower_stmt): Likewise.
7051         * gimple-pretty-print.c (dump_gimple_call): Likewise.
7052         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7053         Likewise.
7054         * gimple.c (gimple_build_call_from_tree): Likewise.
7055         (gimple_call_builtin_p): Likewise.
7056         (gimple_call_combined_fn): Likewise.
7057         * gimplify.c (gimplify_call_expr): Likewise.
7058         (gimple_boolify): Likewise.
7059         (gimplify_modify_expr): Likewise.
7060         (gimplify_addr_expr): Likewise.
7061         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
7062         * ipa-cp.c (determine_versionability): Likewise.
7063         * ipa-fnsummary.c (compute_fn_summary): Likewise.
7064         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
7065         * ipa-split.c (visit_bb): Likewise.
7066         (split_function): Likewise.
7067         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
7068         * lto-cgraph.c (input_node): Likewise.
7069         * lto-streamer-out.c (write_symbol): Likewise.
7070         * omp-low.c (setjmp_or_longjmp_p): Likewise.
7071         (lower_omp_1): Likewise.
7072         * predict.c (strip_predict_hints): Likewise.
7073         * print-tree.c (print_node): Likewise.
7074         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
7075         * trans-mem.c (is_tm_irrevocable): Likewise.
7076         (is_tm_load): Likewise.
7077         (is_tm_simple_load): Likewise.
7078         (is_tm_store): Likewise.
7079         (is_tm_simple_store): Likewise.
7080         (is_tm_abort): Likewise.
7081         (tm_region_init_1): Likewise.
7082         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
7083         * tree-cfg.c (verify_gimple_call): Likewise.
7084         (move_stmt_r): Likewise.
7085         (stmt_can_terminate_bb_p): Likewise.
7086         * tree-eh.c (lower_eh_constructs_2): Likewise.
7087         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
7088         * tree-inline.c (remap_gimple_stmt): Likewise.
7089         (copy_bb): Likewise.
7090         (estimate_num_insns): Likewise.
7091         (fold_marked_statements): Likewise.
7092         * tree-sra.c (scan_function): Likewise.
7093         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
7094         (optimize_stack_restore): Likewise.
7095         (pass_fold_builtins::execute): Likewise.
7096         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
7097         (mark_all_reaching_defs_necessary_1): Likewise.
7098         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
7099         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
7100         (pass_forwprop::execute): Likewise.
7101         * tree-ssa-loop-im.c (stmt_cost): Likewise.
7102         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
7103         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
7104         * tree-ssa-strlen.c (get_string_length): Likewise.
7105         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
7106         (find_func_aliases_for_call): Likewise.
7107         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
7108         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
7109         * tree-tailcall.c (find_tail_calls): Likewise.
7110         * tree.c (need_assembler_name_p): Likewise.
7111         (free_lang_data_in_decl): Likewise.
7112         (get_call_combined_fn): Likewise.
7113         * ubsan.c (is_ubsan_builtin_p): Likewise.
7114         * varasm.c (incorporeal_function_p): Likewise.
7115         * tree.h (DECL_BUILT_IN): Remove and replace with
7116         fndecl_built_in_p.
7117         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
7118         (fndecl_built_in_p): New.
7120 2018-08-27  Martin Liska  <mliska@suse.cz>
7122         PR tree-optimization/86847
7123         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
7124         Dump also subtree probability.
7125         (switch_decision_tree::do_jump_if_equal): New function.
7126         (switch_decision_tree::emit_case_nodes): Handle special
7127         situations in balanced tree that can be emitted much simpler.
7128         Fix calculation of probabilities that happen in tree expansion.
7129         * tree-switch-conversion.h (struct cluster): Add
7130         is_single_value_p.
7131         (struct simple_cluster): Likewise.
7132         (struct case_tree_node): Add new function has_child.
7133         (do_jump_if_equal): New.
7135 2018-08-27  Martin Liska  <mliska@suse.cz>
7137         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
7138         Add new argument to bit_test_cluster constructor.
7139         (bit_test_cluster::emit): Set bits really number of values
7140         handlel by a test.
7141         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
7142         probability argument.
7143         * tree-switch-conversion.h (struct bit_test_cluster):
7144         Add m_handles_entire_switch.
7146 2018-08-27  Martin Liska  <mliska@suse.cz>
7148         PR tree-optimization/86702
7149         * tree-switch-conversion.c (jump_table_cluster::emit):
7150         Make probabilities even for values in jump table
7151         according to number of cases handled.
7152         (switch_decision_tree::compute_cases_per_edge): Pass
7153         argument to reset_out_edges_aux function.
7154         (switch_decision_tree::analyze_switch_statement): Likewise.
7155         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
7156         Make it static.
7158 2018-08-27  Martin Liska  <mliska@suse.cz>
7160         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
7161         cfun argument explicitly.
7162         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
7163         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
7164         function gimple_switch_default_bb.
7165         (convert_switch_statements):
7166         (expand_builtins):
7167         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
7168         * stmt.c (label_to_block_fn): Use label_to_block and pass
7169         cfun argument explicitly and use gimple_switch_label_bb.
7170         (expand_case): Likewise.
7171         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
7172         cfun argument explicitly. Likewise.
7173         (make_edges_bb): Likewise.
7174         (make_cond_expr_edges): Likewise.
7175         (get_cases_for_edge): Likewise.
7176         (make_gimple_switch_edges): Likewise.
7177         (label_to_block_fn): Likewise.
7178         (label_to_block): Likewise.
7179         (make_goto_expr_edges): Likewise.
7180         (make_gimple_asm_edges): Likewise.
7181         (main_block_label): Likewise.
7182         (group_case_labels_stmt): Likewise.
7183         (find_taken_edge_computed_goto): Likewise.
7184         (find_taken_edge_switch_expr): Likewise.
7185         (gimple_verify_flow_info): Likewise.
7186         (gimple_redirect_edge_and_branch): Likewise.
7187         (gimple_switch_label_bb): New function.
7188         (gimple_switch_default_bb): Likewise.
7189         (gimple_switch_edge): Likewise.
7190         (gimple_switch_default_edge): Likewise.
7191         * tree-cfg.h (label_to_block_fn): Remove and replace ...
7192         (label_to_block): ... with this.
7193         (gimple_switch_label_bb): New.
7194         (gimple_switch_default_bb): Likewise.
7195         (gimple_switch_edge): Likewise.
7196         (gimple_switch_default_edge): Likewise.
7197         * tree-cfgcleanup.c (convert_single_case_switch): Use
7198         new gimple functions and pass new argument to label_to_block.
7199         (cleanup_control_flow_bb):
7200         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
7201         cfun argument explicitly.
7202         (make_eh_edges): Likewise.
7203         (redirect_eh_dispatch_edge): Likewise.
7204         (lower_resx): Likewise.
7205         (lower_eh_dispatch): Likewise.
7206         (maybe_remove_unreachable_handlers): Likewise.
7207         (unsplit_eh): Likewise.
7208         (cleanup_empty_eh): Likewise.
7209         (verify_eh_edges): Likewise.
7210         (verify_eh_dispatch_edge): Likewise.
7211         * tree-ssa-dom.c (record_edge_info): Likewise.
7212         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
7213         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
7214         (thread_through_normal_block): Likewise.
7215         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
7216         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
7217         * tree-switch-conversion.c (switch_conversion::collect): Use new
7218         gimple functions.
7219         (switch_conversion::check_final_bb): Likewise.
7220         (switch_conversion::gather_default_values): Pass new argument
7221         to label_to_block.
7222         (switch_conversion::build_constructors): Likewise.
7223         (switch_decision_tree::compute_cases_per_edge): Use new
7224         gimple_switch_edge function.
7225         (switch_decision_tree::analyze_switch_statement): Pass new argument
7226         to label_to_block.
7227         (switch_decision_tree::try_switch_expansion): Use
7228         gimple_switch_default_edge.
7229         * tree-vrp.c (find_switch_asserts): Pass new argument
7230         to label_to_block.
7231         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
7232         (vr_values::simplify_switch_using_ranges): Likewise.
7234 2018-08-27  Richard Biener  <rguenther@suse.de>
7236         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
7237         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
7239         * tree-ssa-sccvn.h (struct vn_pval): New structure.
7240         (struct vn_nary_op_s): Add unwind_to member.  Add
7241         predicated_values flag and put result into a union together
7242         with a linked list of vn_pval.
7243         (struct vn_ssa_aux): Add name member to make maintaining
7244         a map of SSA name to vn_ssa_aux possible.  Remove no longer
7245         needed info, dfsnum, low, visited, on_sccstack, use_processed
7246         and range_info_anti_range_p members.
7247         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
7248         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
7249         New functions.
7250         (vn_valueize): New global.
7251         (vn_context_bb): Likewise.
7252         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
7253         VN_INFO_PTR_INFO): Remove.
7254         * tree-ssa-sccvn.c: ... (rewrite)
7255         (pass_fre::execute): For -O2+ initialize loops and run
7256         RPO VN in optimistic mode (iterating).  For -O1 and -Og
7257         run RPO VN in non-optimistic mode.
7258         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
7259         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
7260         * doc/invoke.texi (sccvn-max-scc-size): Remove.
7261         (rpo-vn-max-loop-depth): Document.
7262         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
7263         when valuezing the VUSE signals we walked out of the region.
7264         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
7265         (phi_translate): Set VN context block to use for availability
7266         lookup.
7267         (compute_avail): Likewise.
7268         (pre_valueize): New function.
7269         (pass_pre::execute): Adjust to the RPO VN API.
7271         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
7272         (propagate_constants_for_unrolling): Remove.
7273         (tree_unroll_loops_completely): Perform value-numbering
7274         on the unrolled bodies loop parent.
7276 2018-08-27  Richard Biener  <rguenther@suse.de>
7278         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
7279         for partial antic compute.
7281 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
7283         PR rtl-optimization/87065
7284         * combine.c (simplify_if_then_else): Formatting fix.
7285         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
7286         check.
7287         (known_cond): Don't return const_true_rtx for vector modes.  Use
7288         CONST0_RTX instead of const0_rtx.  Formatting fixes.
7290 2018-08-27  Martin Liska  <mliska@suse.cz>
7292         PR gcov-profile/87069
7293         * gcov.c (process_file): Record files already processed
7294         and warn about a file being processed multiple times.
7296 2018-08-27  Martin Liska  <mliska@suse.cz>
7298         PR driver/83193
7299         * config/aarch64/aarch64.c (aarch64_override_options_internal):
7300         Set default values for x_aarch64_*_string strings.
7301         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
7302         prefix.  For -mabi do not print '=ABI' in help and use
7303         <option_value> format for -msve-vector-bits and -moverride
7304         options.
7306 2018-08-26  Jeff Law  <law@redhat.com>
7308         * config/mips/frame-header-opt.c: Include "backend.h" rather than
7309         "cfg.h"
7311 2018-08-26  Marek Polacek  <polacek@redhat.com>
7313         PR c++/87029, Implement -Wredundant-move.
7314         * doc/invoke.texi: Document -Wredundant-move.
7316 2018-08-25  Martin Sebor  <msebor@redhat.com>
7318         PR tree-optimization/87059
7319         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
7320         to the same type as the other.
7321         * fold-const.c (fold_binary_loc): Assert expectation.
7323 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
7325         * config/darwin.c (machopic_legitimize_pic_address): Clean up
7326         extraneous parentheses, dead code section and formatting.
7328 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
7330         PR c++/87091
7331         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
7332         wide enough for jumps in the line-numbering to be visible.
7333         (layout::print_gap_in_line_numbering): New member function.
7334         (layout::calculate_line_spans): When using line numbering, merge
7335         line spans that are only 1 line apart.
7336         (diagnostic_show_locus): When printing line numbers, show gaps in
7337         line numbering directly, rather than printing headers.
7338         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
7339         line-numbering with multiple line spans.
7340         (selftest::test_fixit_insert_containing_newline_2): Add test of
7341         line-numbering, in which the spans are close enough to be merged.
7343 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
7345         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
7346         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
7347         * tree-vrp.c (range_is_nonnull): Remove.
7348         (range_includes_zero_p): Accept value_range instead of min/max.
7349         (extract_range_from_binary_expr_1): Do not early bail on
7350         POINTER_PLUS_EXPR.
7351         Use range_includes_zero_p instead of range_is_nonnull.
7352         (extract_range_from_unary_expr): Use range_includes_zero_p instead
7353         of range_is_nonnull.
7354         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
7355         special case VR_ANTI_RANGE.
7356         (vrp_finalize): Same.
7357         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
7358         instead of min/max.
7359         (range_is_nonnull): Remove.
7360         * vr-values.c (vrp_stmt_computes_nonzero): Use
7361         range_includes_zero_p instead of range_is_nonnull.
7362         (extract_range_basic): Pass value_range to range_includes_zero_p
7363         instead of range_is_nonnull.
7365 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
7367         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
7368         * emit-rtl.h (rtl_data): Remove return_bnd.
7369         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
7370         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
7371         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
7372         (POINTER_BOUNDS_MODE): Remove definition.
7373         (make_pointer_bounds_mode): Remove.
7374         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
7375         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
7376         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
7377         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
7378         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
7379         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
7380         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
7382         * config/i386/i386-modes.def (BND32, BND64): Remove.
7383         * config/i386/i386.c (dbx_register_map): Remove bound registers.
7384         (dbx64_register_map): Ditto.
7385         (svr4_dbx_register_map): Ditto.
7386         (indirect_thunk_bnd_needed): Remove.
7387         (indirect_thunks_bnd_used): Ditto.
7388         (indirect_return_bnd_needed): Ditto.
7389         (indirect_return_via_cx_bnd): Ditto.
7390         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
7391         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
7392         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
7393         (output_indirect_thunk_function): Remove handling of
7394         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
7395         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
7396         (ix86_save_reg): Remove handling of crtl->return_bnd.
7397         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
7398         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
7399         and UNSPEC_BNDLX_ADDR.
7400         (ix86_output_indirect_branch_via_reg): Remove handling of
7401         indirect_thunk_prefix_bnd.
7402         (ix86_output_indirect_branch_via_push): Ditto.
7403         (ix86_output_function_return): Ditto.
7404         (ix86_output_indirect_function_return): Ditto.
7405         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
7406         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
7407         (CALL_USED_REGISTERS): Ditto.
7408         (REG_ALLOC_ORDER): Update for removal of bound registers.
7409         (HI_REGISTER_NAMES): Ditto.
7410         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
7411         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
7412         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
7413         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
7414         (FIRST_PSEUDO_REG): Update.
7415         (BND): Remove mode iterator.
7416         * config/i386/predicates.md (bnd_mem_operator): Remove.
7418 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
7420         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
7421         vectors.
7423 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
7425         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
7426         the case in which the permute needs only a single element and
7427         repeats for every vector of the result.  Extend that case to
7428         handle variable-length vectors.
7429         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
7431 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
7433         PR debug/79342
7434         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
7435         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
7437 2018-08-24  Richard Biener  <rguenther@suse.de>
7439         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
7440         bb_flags_allocated members.
7441         (auto_flag): New RAII class for allocating flags.
7442         (auto_edge_flag): New RAII class for allocating edge flags.
7443         (auto_bb_flag): New RAII class for allocating bb flags.
7444         * cfgloop.c (verify_loop_structure): Allocate temporary edge
7445         flag dynamically.
7446         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
7447         in favor of temporarily allocated BB flag.
7448         * hsa-brig.c: Re-order includes.
7449         * hsa-dump.c: Likewise.
7450         * hsa-regalloc.c: Likewise.
7451         * print-rtl.c: Likewise.
7452         * profile-count.c: Likewise.
7454 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
7456         PR target/86989
7457         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
7458         the TOC register.
7460 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
7462         PR 87073/bootstrap
7463         * wide-int-range.cc (wide_int_range_div): Do not ignore result
7464         from wide_int_range_multiplicative_op.
7466 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
7468         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
7469         "permutaion".
7471 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
7473         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
7474         to 'expanded'.
7476 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
7478         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
7479         full GENERIC expression used for replacement.
7481 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
7483         * tree-vrp.c (abs_extent_range): Remove.
7484         (extract_range_into_wide_ints): Pass wide ints by reference.
7485         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
7486         Pass wide ints by reference in all calls to
7487         extract_range_into_wide_ints.
7488         * wide-int-range.cc (wide_int_range_div): New.
7489         * wide-int-range.h (wide_int_range_div): New.
7490         (wide_int_range_includes_zero_p): New.
7491         (wide_int_range_zero_p): New.
7493 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
7495         * config/aarch64/aarch64.md (arches): New enum.
7496         (arch): New enum attr.
7497         (arch_enabled): New attr.
7498         (enabled): Now uses arch_enabled only.
7499         (simd, sve, fp16): Removed attribute.
7500         (fp): Attr now defined in terms of 'arch'.
7501         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
7502         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
7503         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
7504         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
7505         attributes into 'arch'.
7506         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
7507         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
7508         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
7509         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
7510         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
7511         'simd' attribute into 'arch'.
7512         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
7513         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
7514         Convert use of 'fp' attribute to 'arch'.
7515         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
7516         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
7517         into 'arch'.
7518         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
7519         (different modes) Merge 'fp' and 'simd' into 'arch'.
7520         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
7521         'simd' into 'arch'.
7523 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
7525         PR rtl-optimization/87026
7526         * expmed.c (canonicalize_comparison): If we can no longer create
7527         pseudoregisters, don't.
7529 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
7531         PR target/86951
7532         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
7533         prototype.
7534         * config/arm/arm.c (speculation_barrier_libfunc): New static
7535         variable.
7536         (arm_init_libfuncs): Initialize it.
7537         (arm_emit_speculation_barrier): New function.
7538         * config/arm/arm.md (speculation_barrier): Call
7539         arm_emit_speculation_barrier for architectures that do not have
7540         DSB or ISB.
7541         (speculation_barrier_insn): Only match on Armv7 or later.
7543 2018-08-23  Richard Biener  <rguenther@suse.de>
7545         PR middle-end/87024
7546         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
7547         calls.
7549 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
7551         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
7552         of single-vector TBLs.
7553         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
7554         one input is given.
7556 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
7558         PR target/85910
7559         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
7560         aarch64_evpc_tbl guard.
7562 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7564         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
7565         behaviour.
7567 2018-08-22  Martin Sebor  <msebor@redhat.com>
7569         PR middle-end/87052
7570         * tree-pretty-print.c (pretty_print_string): Add argument.
7571         (dump_generic_node): Call to pretty_print_string with string size.
7573 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
7575         PR rtl-optimization/86771
7576         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
7577         of two SETs into those two SETs, one to be placed at i2, if that SETs
7578         destination is modified between i2 and i3.
7580 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
7582         PR tree-optimization/86725
7583         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
7584         function.
7585         (vect_analyze_scalar_cycles_1): Check it.
7587 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
7589         PR tree-optimization/86725
7590         * tree-vect-loop.c (vect_is_simple_reduction): When treating
7591         an outer loop phi as a double reduction, make sure that the
7592         single user of the phi result is an inner loop phi.
7594 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
7596         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
7597         grouped stores with gaps to a strided group.
7599 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
7601         * tree-vect-stmts.c (get_group_load_store_type)
7602         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
7603         first statement in a group.
7605 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
7607         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
7608         the sequence used in gcc/gcc.c.
7610 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
7612         PR other/704
7613         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
7614         building it.
7616 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
7618         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
7619         Darwin10-specific unwinder-shim.
7620         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
7621         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
7622         New to cater for Darwin10 Rosetta.
7624 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
7626         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
7627         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
7629 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
7631         PR bootstrap/81033
7632         PR target/81733
7633         PR target/52795
7634         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
7635         (dwarf2out_switch_text_section): Generate a local label for the
7636         second function sub-section and apply it as the second FDE start
7637         label.
7638         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
7639         second sub-section start.
7641 2018-08-22  Richard Biener  <rguenther@suse.de>
7643         PR tree-optimization/86988
7644         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
7646 2018-08-22  Richard Biener  <rguenther@suse.de>
7648         PR tree-optimization/86945
7649         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
7651 2018-08-22  Alexandre Oliva <oliva@adacore.com>
7653         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
7654         a comment about how uses of r2 for .sdata2 come about.
7656 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
7658         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
7660 2018-08-21  Marek Polacek  <polacek@redhat.com>
7662         PR c++/86981, Implement -Wpessimizing-move.
7663         * doc/invoke.texi: Document -Wpessimizing-move.
7665 2018-08-21  Jan Hubicka  <jh@suse.cz>
7667         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
7668         * tree.h (is_redundant_typedef): Remove.
7669         * dwarf2out.c (is_redundant_typedef): Turn into static function.
7671 2018-08-21  Jan Hubicka  <jh@suse.cz>
7673         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
7674         when possible.
7676 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
7678         * expmed.c (extract_low_bits): Reject invalid subregs early.
7680 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7682         PR middle-end/86121
7683         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
7684         behaviour.
7686 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
7688         * config/vxworks.h: Guard vxworks_asm_out_constructor and
7689         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
7690         * config/vxworks.c: Likewise.
7692 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
7694         * config/vxworks.c: Set targetm.have_ctors_dtors
7695         if HAVE_INITFINI_ARRAY_SUPPORT.
7696         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
7697         if HAVE_INITFINI_ARRAY_SUPPORT.
7699 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
7701         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
7702         default search path for VxWorks < 7.
7704 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7706         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
7707         (get_format_string): Refer to c_getstr.
7709 2018-08-21  Tom de Vries  <tdevries@suse.de>
7711         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
7712         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
7713         (debuginfo_early_stop): Declare.
7714         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
7715         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
7716         (debuginfo_early_stop): New function.
7717         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
7718         and debuginfo_early_stop.
7719         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
7720         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
7721         (general_init): Call debuginfo_early_init.
7722         (finalize): Call debuginfo_fini.
7723         (do_compile): Call debuginfo_init.
7724         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
7725         -fdump-early-debug.
7726         (@item -fdump-debug, @item -fdump-earlydebug): Add.
7728 2018-08-21  Tom de Vries  <tdevries@suse.de>
7730         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
7731         flag_dump_noaddr and flag_dump_unnumbered.
7733 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
7735         * wide-int-range.cc (wide_int_range_abs): New.
7736         (wide_int_range_order_set): Rename from wide_int_range_min_max.
7737         * wide-int-range.h (wide_int_range_abs): New.
7738         (wide_int_range_min_max): New.
7739         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
7740         case to call wide_int_range_abs.
7741         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
7742         (extract_range_from_abs_expr): Delete.
7744 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
7746         PR target/87033
7747         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
7748         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
7749         for indexed loads.
7751 2018-08-20  Nathan Sidwell  <nathan@acm.org>
7752             Jeff Law <law@redhat.com>
7754         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
7755         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
7757 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
7759         PR other/84889
7760         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
7761         (decl_attributes): Likewise.
7762         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
7763         instance.
7764         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
7765         * diagnostic-core.h (class auto_diagnostic_group): New class.
7766         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
7767         (diagnostic_report_diagnostic): Handle the first diagnostics within
7768         a group.
7769         (emit_diagnostic): Add auto_diagnostic_group instance.
7770         (inform): Likewise.
7771         (inform_n): Likewise.
7772         (warning): Likewise.
7773         (warning_at): Likewise.
7774         (warning_n): Likewise.
7775         (pedwarn): Likewise.
7776         (permerror): Likewise.
7777         (error): Likewise.
7778         (error_n): Likewise.
7779         (error_at): Likewise.
7780         (sorry): Likewise.
7781         (fatal_error): Likewise.
7782         (internal_error): Likewise.
7783         (internal_error_no_backtrace): Likewise.
7784         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
7785         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
7786         * diagnostic.h (struct diagnostic_context): Add fields
7787         "diagnostic_group_nesting_depth",
7788         "diagnostic_group_emission_count", "begin_group_cb",
7789         "end_group_cb".
7790         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
7791         Add auto_diagnostic_group instance(s).
7792         (find_explicit_erroneous_behavior): Likewise.
7793         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
7794         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
7795         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
7796         (gimplify_va_arg_expr): Likewise.
7797         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
7798         (HSA_SORRY_AT): Likewise.
7799         * ipa-devirt.c (compare_virtual_tables): Likewise.
7800         (warn_odr): Likewise.
7801         * multiple_target.c (expand_target_clones): Likewise.
7802         * opts-common.c (cmdline_handle_error): Likewise.
7803         * reginfo.c (globalize_reg): Likewise.
7804         * substring-locations.c (format_warning_n_va): Likewise.
7805         * tree-inline.c (expand_call_inline): Likewise.
7806         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
7807         * tree-ssa-loop-niter.c
7808         (do_warn_aggressive_loop_optimizations): Likewise.
7809         * tree-ssa-uninit.c (warn_uninit): Likewise.
7810         * tree.c (warn_deprecated_use): Likewise.
7812 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
7814         PR target/87014
7815         * config/i386/i386.md (eh_return): Always update EH return
7816         address in word_mode.
7818 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
7820         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
7821         TARGET_SPLIT_COMPLEX_ARG is defined.
7823 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7825         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
7827 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7829         PR target/86984
7830         * expr.c (expand_assignment): Assert that bitpos is positive.
7831         (store_field): Likewise
7832         (expand_expr_real_1): Make sure that bitpos is positive.
7833         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
7834         integer overflow.
7836 2018-08-20  Nathan Sidwell  <nathan@acm.org>
7838         * Makefile.in (CPP_ID_DATA_H): Delete.
7839         (CPP_INTERNAL_H): Don't add it.
7840         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
7841         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
7843 2018-08-20  Richard Biener  <rguenther@suse.de>
7845         PR tree-optimization/78655
7846         * tree-vrp.c (extract_range_from_binary_expr_1): Make
7847         pointer + offset nonnull if either operand is nonnull work.
7849 2018-08-20  Tom de Vries  <tdevries@suse.de>
7851         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
7852         unless the referenced die describes the added property using
7853         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
7854         Otherwise, add a DW_AT_location to the referenced die.
7856 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
7858         PR target/86994
7859         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
7860         register_operand when calling ix86_set_reg_reg_cost.
7861         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
7862         Set *total to 0 for operands that satisfy x86_64_immediate_operand
7863         predicate and to 1 otherwise.
7865 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
7867         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
7868         emit a diagnostic that it is not supported and reset the option.
7869         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
7870         supported and consume the option.  (ASM_FINAL_SPEC): New.
7872 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
7874         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
7875         a sentence.
7877 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
7879         C-SKY port: Documentation
7881         * doc/extend.texi (C-SKY Function Attributes): New section.
7882         * doc/invoke.texi (Option Summary): Add C-SKY options.
7883         (C-SKY Options): New section.
7884         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
7886 2018-08-17  Jojo  <jijie_rong@c-sky.com>
7887             Huibin Wang  <huibin_wang@c-sky.com>
7888             Sandra Loosemore  <sandra@codesourcery.com>
7889             Chung-Lin Tang  <cltang@codesourcery.com>
7891         C-SKY port: Backend implementation
7893         * config/csky/*: New.
7894         * common/config/csky/*: New.
7896 2018-08-17  Jojo  <jijie_rong@c-sky.com>
7897             Huibin Wang  <huibin_wang@c-sky.com>
7898             Sandra Loosemore  <sandra@codesourcery.com>
7899             Chung-Lin Tang  <cltang@codesourcery.com>
7900             Andrew Jenner  <andrew@codesourcery.com>
7902         C-SKY port: Configury
7904         * config.gcc (csky-*-*): New.
7905         * configure.ac: Add csky to targets for dwarf2 debug_line support.
7906         * configure: Regenerated.
7908 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
7910         * dump-context.h: Include "dumpfile.h".
7911         (dump_context::dump_printf_va): Convert final param from va_list
7912         to va_list *.  Convert from ATTRIBUTE_PRINTF to
7913         ATTRIBUTE_GCC_DUMP_PRINTF.
7914         (dump_context::dump_printf_loc_va): Likewise.
7915         * dumpfile.c: Include "stringpool.h".
7916         (make_item_for_dump_printf_va): Delete.
7917         (make_item_for_dump_printf): Delete.
7918         (class dump_pretty_printer): New class.
7919         (dump_pretty_printer::dump_pretty_printer): New ctor.
7920         (dump_pretty_printer::emit_items): New member function.
7921         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
7922         function.
7923         (dump_pretty_printer::emit_item): New member function.
7924         (dump_pretty_printer::stash_item): New member function.
7925         (dump_pretty_printer::format_decoder_cb): New member function.
7926         (dump_pretty_printer::decode_format): New member function.
7927         (dump_context::dump_printf_va): Reimplement in terms of
7928         dump_pretty_printer.
7929         (dump_context::dump_printf_loc_va): Convert final param from va_list
7930         to va_list *.
7931         (dump_context::begin_scope): Reimplement call to
7932         make_item_for_dump_printf.
7933         (dump_printf): Update for change to dump_printf_va.
7934         (dump_printf_loc): Likewise.
7935         (selftest::test_capture_of_dump_calls): Convert "stmt" from
7936         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
7937         with %T, %E, and %G.
7938         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
7939         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
7940         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
7941         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
7942         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
7943         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
7944         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
7945         within a dump_printf_loc call to "%wu".
7946         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
7947         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
7948         missing space after "=".
7949         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
7950         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
7951         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
7952         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
7953         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
7954         duplicate "vectorized" from message.
7956 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
7958         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
7959         polyNxK_t element's TYPE_STRING_FLAG.
7961 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
7963         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
7964         (they were unnamed before).  Fix comments.
7966 2018-08-17  Nathan Sidwell  <nathan@acm.org>
7968         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
7970 2018-08-17  Richard Biener  <rguenther@suse.de>
7972         PR tree-optimization/86841
7973         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
7975 2018-08-17  Martin Liska  <mliska@suse.cz>
7977         * common.opt: Remove Warn, Init and Report for options with
7978         Ignore/Deprecated flag. Warning is done automatically for
7979         Deprecated flags.
7980         * config/i386/i386.opt: Likewise.
7981         * config/ia64/ia64.opt: Likewise.
7982         * config/rs6000/rs6000.opt: Likewise.
7983         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
7984         Remove usage of flag_check_pointer_bounds.
7985         * lto-wrapper.c (merge_and_complain): Do not handle
7986         OPT_fcheck_pointer_bounds.
7987         (append_compiler_options): Likewise.
7988         * opt-functions.awk: Do not handle Deprecated.
7989         * optc-gen.awk: Check that Var, Report and Init are not
7990         used for an option with Ignore/Deprecated flag.
7991         * opts-common.c (decode_cmdline_option): Do not report
7992         CL_ERR_DEPRECATED.
7993         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
7994         options.
7995         * opts.h (struct cl_option): Remove cl_deprecated flag.
7996         (CL_ERR_DEPRECATED): Remove error enum value.
7998 2018-08-17  Richard Biener  <rguenther@suse.de>
8000         PR middle-end/86505
8001         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
8002         across a va-arg-pack using call adjust its return value accordingly.
8004 2018-08-16  Martin Sebor  <msebor@redhat.com>
8006         PR tree-optimization/86853
8007         * gimple-ssa-sprintf.c (struct format_result): Rename member.
8008         (struct fmtresult): Add member and initialize it in ctors.
8009         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
8010         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
8011         (format_directive): Set POSUNDER4K when MAYFAIL is set.
8012         (parse_directive): Handle %C same as %c.
8013         (sprintf_dom_walker::compute_format_length): Adjust.
8014         (is_call_safe): Adjust.
8016 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8018         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
8019         for determining how to count the elements.
8020         * builtins.h (c_strlen): Adjust prototype.
8021         * expr.c (string_constant): Add new parameter mem_size.
8022         Set *mem_size appropriately.
8023         * expr.h (string_constant): Adjust protoype.
8024         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
8025         * gimple-fold.h (get_range_strlen): Adjust prototype.
8026         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
8027         (format_string): Call get_string_length with eltsize.
8029 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
8031         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
8032         to emit the span, rather than setting it as the prefix.
8034 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
8036         * diagnostic-show-locus.c (layout::start_annotation_line): Add
8037         "margin_char" parameter, defaulting to space.  Use it in place
8038         of pp_space for the initial part of the margin.
8039         (layout::print_leading_fixits): Use '+' when filling the margin
8040         of line-insertion fix-it hints.
8042 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8044         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
8045         Delete.
8047 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8049         * config/rs6000/altivec.md: Don't set length attribute to the default
8050         value.
8051         * config/rs6000/darwin.md: Ditto.
8052         * config/rs6000/dfp.md: Ditto.
8053         * config/rs6000/htm.md: Ditto.
8054         * config/rs6000/rs6000.md: Ditto.
8055         * config/rs6000/sync.md: Ditto.
8056         * config/rs6000/vsx.md: Ditto.
8058 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8060         * config/rs6000/altivec.md: Don't set length attribute to the default
8061         value, for branch instructions.
8062         * config/rs6000/darwin.md: Ditto.
8063         * config/rs6000/rs6000.md: Ditto.
8065 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
8067         * config/rs6000/rs6000.md (length): Always define as const_int 4.
8068         (unnamed conditional branch define_insn): Set length to 4 or 8
8069         depending on offset.
8070         (<bd>_<mode>): Similar, for alternative 0.
8071         (<bd>tf_<mode>): Ditto.
8073 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
8075         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
8077 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
8079         * doc/rtl.texi: Replace old RTX class names with new names.
8082 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
8084         * expmed.h (canonicalize_comparison): New declaration.
8085         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
8086         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
8087         * optabs.c (prepare_cmp_insn): Likewise.
8088         * rtl.h (unsigned_condition_p): New function which checks if a
8089         comparison operator is unsigned.
8091 2018-08-16  Nathan Sidwell  <nathan@acm.org>
8093         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
8094         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
8096 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
8098         PR target/84711
8099         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
8100         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
8101         (mov<mov>): ..this and enable unconditionally.
8103 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
8105         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
8107 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
8109         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
8110         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
8111         "Common" with "Target".
8113 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
8115         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
8116         * doc/invoke.texi (mmitigate-rop): Remove.
8117         * config/i386/i386.c: Do not include "regrename.h".
8118         (ix86_rop_should_change_byte_p, reg_encoded_number)
8119         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
8120         Remove.
8121         (ix86_reorg): Remove call to ix86_mitigate_rop.
8122         * config/i386/i386.md (attr "modrm_class"): Remove.
8123         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
8124         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
8125         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
8127 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
8129         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
8130         allow folding of mergeh() and mergel() for the float and double types.
8131         (fold_mergehl_helper): Rework to handle building a permute tree
8132         for float vectors.
8134 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
8136         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
8137         for TARGET_SSE.
8139 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
8141         * common.opt (fdiagnostics-show-labels): New option.
8142         * diagnostic-show-locus.c (class layout_range): Add field
8143         "m_label".
8144         (class layout): Add field "m_show_labels_p".
8145         (layout_range::layout_range): Add param "label" and use it to
8146         initialize m_label.
8147         (make_range): Pass in NULL for new "label" param of layout_range's
8148         ctor.
8149         (layout::layout): Initialize m_show_labels_p.
8150         (layout::maybe_add_location_range): Pass in loc_range->m_label
8151         when constructing layout_range instances.
8152         (struct line_label): New struct.
8153         (layout::print_any_labels): New member function.
8154         (layout::print_line): Call it if label-printing is enabled.
8155         (selftest::test_one_liner_labels): New test.
8156         (selftest::test_diagnostic_show_locus_one_liner): Call it.
8157         * diagnostic.c (diagnostic_initialize): Initialize
8158         context->show_labels_p.
8159         * diagnostic.h (struct diagnostic_context): Add field
8160         "show_labels_p".
8161         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
8162         -fno-diagnostics-show-labels.
8163         * dwarf2out.c (gen_producer_string): Add
8164         OPT_fdiagnostics_show_labels to the ignored options.
8165         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
8166         param.
8167         (gcc_rich_location::maybe_add_expr): Likewise.
8168         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
8169         label" param, defaulting to NULL.
8170         (gcc_rich_location::add_expr): Add "label" param.
8171         (gcc_rich_location::maybe_add_expr): Likewise.
8172         (class text_range_label): New class.
8173         (class range_label_for_type_mismatch): New class.
8174         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
8175         of format_warning_va.
8176         (fmtwarn_n): Likewise for new params of format_warning_n_va.
8177         * lto-wrapper.c (merge_and_complain): Add
8178         OPT_fdiagnostics_show_labels to the "pick one setting" options.
8179         (append_compiler_options): Likewise to the dropped options.
8180         (append_diag_options): Likewise to the passed-on options.
8181         * opts.c (common_handle_option): Handle the new option.
8182         * selftest-diagnostic.c
8183         (test_diagnostic_context::test_diagnostic_context): Enable
8184         show_labels_p.
8185         * substring-locations.c: Include "gcc-rich-location.h".
8186         (format_warning_n_va): Add "fmt_label" and "param_label" params
8187         and use them as appropriate.
8188         (format_warning_va): Add "fmt_label" and "param_label" params,
8189         passing them on to format_warning_n_va.
8190         (format_warning_at_substring): Likewise.
8191         (format_warning_at_substring_n): Likewise.
8192         * substring-locations.h (format_warning_va): Add "fmt_label" and
8193         "param_label" params.
8194         (format_warning_n_va): Likewise.
8195         (format_warning_at_substring): Likewise.
8196         (format_warning_at_substring_n): Likewise.
8197         * toplev.c (general_init): Initialize global_dc->show_labels_p.
8199 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
8201         PR testsuite/86519
8202         * builtins.c (expand_builtin_memcmp): Do not expand the call
8203         when overflow is detected.
8205 2018-08-15  Martin Sebor  <msebor@redhat.com>
8207         PR tree-optimization/71625
8208         * config/aarch64/aarch64-builtins.c
8209         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
8211 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
8213         * config/s390/s390.c (s390_reorg): Remove loop.
8215 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
8217         * config/darwin.c
8218          (darwin_function_switched_text_sections): Delete.
8219         * gcc/config/darwin.h
8220          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
8222 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
8224         PR target/81685
8225         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
8226         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
8227         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
8229 2018-08-15  Martin Liska  <mliska@suse.cz>
8231         PR tree-optimization/86925
8232         * predict.c (expr_expected_value_1): When taking
8233         later predictor, assign also probability.
8234         Use fold_build2_initializer_loc in order to fold
8235         the expression in -frounding-math.
8237 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
8239         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
8240         patterns.
8241         (expand_vec_perm_1): Try the new method.
8243 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
8245         PR target/86547
8246         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
8247         Check whether lra_live_max_point is 0 before dividing.
8249 2018-08-14  Martin Sebor  <msebor@redhat.com>
8251         PR tree-optimization/86650
8252         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
8253         (vrp_prop::check_mem_ref): Same.
8255 2018-08-13  Liu Hao <lh_mouse@126.com>
8257         * pretty-print.c (eat_esc_sequence): Swap the foreground and
8258         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
8259         and clear it thereafter, as it only works for DBCS.
8261 2018-08-13  Liu Hao <lh_mouse@126.com>
8263         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
8264         handle returned by _get_osf_handle().
8266 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
8268         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
8269         for folding vec_perm.
8271 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
8273         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
8274         Add support for gimple-folding of vec_pack() and vec_unpack()
8275         intrinsics.
8277 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
8279         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
8280         vec_xst variants to the list.
8281         (rs6000_gimple_fold_builtin): Add support for folding unaligned
8282         vector loads and stores.
8284 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
8286         * config.gcc (rs6000-ibm-aix4.x): Delete.
8287         (rs6000-ibm-aix5.1): Delete.
8288         (rs6000-ibm-aix5.2): Delete.
8289         (rs6000-ibm-aix5.3): Delete.
8290         * config/rs6000/aix43.h: Delete.
8291         * config/rs6000/aix51.h: Delete.
8292         * config/rs6000/aix52.h: Delete.
8293         * config/rs6000/t-aix43: Delete.
8295 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
8297         * config/s390/s390.c (s390_decompose_constant_pool_ref):
8298         New function.
8299         (s390_decompose_address): Factor out constant pool ref
8300         decomposition.
8302 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
8304         * config/nds32/nds32-predicates.c
8305         (nds32_can_use_bclr_p): Change return type as bool.
8306         (nds32_can_use_bset_p): Ditto.
8307         (nds32_can_use_btgl_p): Ditto.
8308         (nds32_can_use_bitci_p): Ditto.
8309         * config/nds32/nds32-protos.h
8310         (nds32_can_use_bclr_p): Change declaration.
8311         (nds32_can_use_bset_p): Ditto.
8312         (nds32_can_use_btgl_p): Ditto.
8313         (nds32_can_use_bitci_p): Ditto.
8315 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
8317         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
8318         Support -msched-prolog-epilog option.
8319         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
8321 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
8323         * common/config/nds32/nds32-common.c
8324         (nds32_option_optimization_table): Enalbe -malways-align.
8326 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
8328         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
8329         extra_headers.
8330         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
8331         OPT_misr_secure_ case.
8332         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
8333         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
8334         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
8335         secure attribute.
8336         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
8337         (nds32_isr_info): New field security_level.
8338         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
8339         * config/nds32/nds32.md (return_internal): Consider critical attribute.
8340         * config/nds32/nds32.opt (misr-secure): New option.
8341         * config/nds32/nds32_init.inc: New file.
8342         * config/nds32/nds32_isr.h: New file.
8344 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
8346         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
8347         Update comment for atomic instructions.
8348         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
8349         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
8350         Remove.
8351         (atomic_loaddi): Revise fence expansion to only emit fence prior to
8352         load for __ATOMIC_SEQ_CST model.
8353         (atomic_loaddi_1): Remove float register target.
8354         (atomic_storedi): Handle CONST_INT values.
8355         (atomic_storedi_1): Remove float register source.  Add special case
8356         for zero value.
8357         (memory_barrier): New expander and insn.
8359 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
8361         PR tree-optimization/86835
8362         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
8363         new_stmt after def_gsi, make sure to insert new_square_stmt after
8364         that stmt, not 2 stmts before it.
8366 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
8368         PR target/82418
8369         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
8370         instead of SWI48.
8372 2018-08-10  Martin Liska  <mliska@suse.cz>
8374         PR target/83610
8375         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
8376         function type.
8377         * builtins.c (expand_builtin_expect_with_probability):
8378         New function.
8379         (expand_builtin_expect_with_probability): New function.
8380         (build_builtin_expect_predicate): Add new argumnet probability
8381         for BUILT_IN_EXPECT_WITH_PROBABILITY.
8382         (fold_builtin_expect):
8383         (fold_builtin_2):
8384         (fold_builtin_3):
8385         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
8386         * builtins.h (fold_builtin_expect): Set new argument.
8387         * doc/extend.texi: Document __builtin_expect_with_probability.
8388         * doc/invoke.texi: Likewise.
8389         * gimple-fold.c (gimple_fold_call): Pass new argument.
8390         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
8391         also BUILT_IN_EXPECT_WITH_PROBABILITY.
8392         * predict.c (get_predictor_value): New function.
8393         (expr_expected_value): Add new argument probability. Assume
8394         that predictor and probability are always non-null.
8395         (expr_expected_value_1): Likewise.  For __builtin_expect and
8396         __builtin_expect_with_probability set probability.  Handle
8397         combination in binary expressions.
8398         (tree_predict_by_opcode): Simplify code by simply calling
8399         get_predictor_value.
8400         (pass_strip_predict_hints::execute): Add handling of
8401         BUILT_IN_EXPECT_WITH_PROBABILITY.
8402         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
8403         new predictor.
8404         * tree.h (DECL_BUILT_IN_P): New function.
8406 2018-08-10  Martin Liska  <mliska@suse.cz>
8408         PR tree-optimization/85799
8409         * passes.def: Add argument for pass_strip_predict_hints.
8410         * predict.c (class pass_strip_predict_hints): Add new argument
8411         early_p.
8412         (strip_predictor_early): New function.
8413         (pass_strip_predict_hints::execute): Call the function to
8414         strip predictors.
8415         (strip_predict_hints): New function.
8416         * predict.def: Fix comment.
8418 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
8420         * Makefile.in: Clarify which tm.texi to copy over to assert the
8421         right to grant a GFDL license for all.
8423 2018-08-09  Jeff Law  <law@redhat.com>
8425         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
8426         unused variable.
8428 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
8430         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
8431         prototype.
8433 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
8435         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
8436         reductions for variable-length vectors.
8438 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
8440         PR other/84889
8441         * common.opt (fdiagnostics-show-line-numbers): New option.
8442         * diagnostic-show-locus.c (class layout): Add fields
8443         "m_show_line_numbers_p" and "m_linenum_width";
8444         (num_digits): New function.
8445         (test_num_digits): New function.
8446         (layout::layout): Initialize new fields.  Update m_x_offset
8447         logic to handle any left margin.
8448         (layout::print_source_line): Print line number when requested.
8449         (layout::start_annotation_line): New member function.
8450         (layout::print_annotation_line): Call it.
8451         (layout::print_leading_fixits): Likewise.
8452         (layout::print_trailing_fixits): Likewise.  Update calls to
8453         move_to_column for new parameter.
8454         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
8455         it to potentially call start_annotation_line.
8456         (layout::show_ruler): Call start_annotation_line.
8457         (selftest::test_line_numbers_multiline_range): New selftest.
8458         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
8459         and selftest::test_line_numbers_multiline_range.
8460         * diagnostic.c (diagnostic_initialize): Initialize
8461         show_line_numbers_p.
8462         * diagnostic.h (struct diagnostic_context): Add field
8463         "show_line_numbers_p".
8464         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
8465         -fno-diagnostics-show-line-numbers.
8466         * dwarf2out.c (gen_producer_string): Add
8467         OPT_fdiagnostics_show_line_numbers to the ignored options.
8468         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
8469         one setting" options.
8470         (append_compiler_options): Likewise to the dropped options.
8471         (append_diag_options): Likewise to the passed-on options.
8472         * opts.c (common_handle_option): Handle the new option.
8473         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
8475 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8477         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
8478         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
8479         third argument of type "const signed char" to descriptions of
8480         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
8481         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
8482         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
8483         __builtin_bcdsub_ov functions.
8485 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
8487         PR tree-optimization/86858
8488         * tree-vect-loop.c (vect_is_simple_reduction): Restore
8489         flow_bb_inside_loop_p calls.
8491 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
8493         PR tree-optimization/86871
8494         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
8495         instead of gimple_assign_lhs.
8497 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
8499         PR target/86887
8500         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
8501         register constraint to operand 0.
8502         (add<mode>3_carryinC): Likewise.
8503         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
8505 2018-08-09  Martin Liska  <mliska@suse.cz>
8507         PR c/86895
8508         * common.opt: Remove extra line.
8510 2018-08-09  Martin Liska  <mliska@suse.cz>
8512         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
8513         at the end of a line, make first letter capital and end up
8514         a sentence with a dot.
8515         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
8516         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
8517         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
8518         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
8519         (PARAM_MAX_ISL_OPERATIONS): Likewise.
8520         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
8521         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
8522         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
8523         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
8524         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
8525         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
8526         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
8527         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
8528         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
8529         (PARAM_TREE_REASSOC_WIDTH): Likewise.
8530         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
8531         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
8532         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
8534 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
8536         PR target/84332
8537         * config/s390/s390.c (s390_option_override_internal): Reduce the
8538         stack-clash-protection-probe-interval param if it would be too big
8539         for z900.
8541 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
8543         PR target/46179
8544         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
8545         * config/m68k/m68k.c (handle_move_double): Don't call
8546         m68k_final_prescan_insn.
8547         (m68k_adjust_decorated_operand): Renamed from
8548         m68k_final_prescan_insn, remove first and third operand and
8549         simplify.
8550         (print_operand): Call it.
8551         (print_operand_address): Call it.
8553 2018-08-08  Nathan Sidwell  <nathan@acm.org>
8555         * diagnostic.c (diagnostic_report_current_module): Use
8556         linemap_included_from & linemap_included_from_linemap.
8558 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
8560         * config/aarch64/aarch64-cores.def: Add phecda core.
8561         * config/aarch64/aarch64-tune.md: Regenerate.
8562         * doc/invoke.texi: Add phecda core.
8564 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
8566         PR target/85295
8567         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
8568         definitions.
8569         * config/s390/s390.md ("movti"): Add more alternatives for
8570         constant to GPR copies.
8572 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
8574         * config/s390/s390.c: Fix whitespace damage throughout the file.
8575         * config/s390/s390.h: Likewise.
8576         * config/s390/tpf.h: Likewise.
8578 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
8580         * config/s390/s390.c (s390_loadrelative_operand_p):
8581         Remove TARGET_CPU_ZARCH usages.
8582         (s390_rtx_costs): Likewise.
8583         (s390_legitimate_constant_p): Likewise.
8584         (s390_cannot_force_const_mem): Likewise.
8585         (legitimate_reload_constant_p): Likewise.
8586         (s390_preferred_reload_class): Likewise.
8587         (legitimize_pic_address): Likewise.
8588         (legitimize_tls_address): Likewise.
8589         (s390_split_branches): Removed.
8590         (s390_add_execute): Removed.
8591         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
8592         (s390_mainpool_start): Likewise.
8593         (s390_mainpool_finish): Likewise.
8594         (s390_mainpool_cancel): Removed.
8595         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
8596         (s390_chunkify_cancel): Likewise.
8597         (s390_return_addr_rtx): Likewise.
8598         (s390_register_info): Remove split_branches_pending_p uages.
8599         (s390_optimize_register_info): Likewise.
8600         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
8601         split_branches_pending_p usages.
8602         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
8603         (s390_load_got): Likewise.
8604         (s390_expand_split_stack_prologue): Likewise.
8605         (output_asm_nops): Likewise.
8606         (s390_function_profiler): Likewise.
8607         (s390_emit_call): Likewise.
8608         (s390_conditional_register_usage): Likewise.
8609         (s390_optimize_prologue): Likewise.
8610         (s390_reorg): Remove TARGET_CPU_ZARCH and
8611         split_branches_pending_p usages.
8612         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
8613         usages.
8614         (s390_output_indirect_thunk_function): Likewise.
8615         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
8616         (TARGET_CPU_ZARCH_P): Removed.
8617         (struct machine_function): Remove split_branches_pending_p.
8618         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
8620 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
8622         * common/config/s390/s390-common.c (processor_flags_table):
8623         Remove flags.
8624         * config.gcc: Remove with_arch/with_tune support.
8625         * config/s390/2064.md: Remove cpu attribute comparisons.
8626         * config/s390/driver-native.c (s390_host_detect_local_cpu):
8627         Remove MTN.
8628         * config/s390/linux.h (ASM_SPEC):
8629         Remove -march support.
8630         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
8631         Use a table to get an arch level.
8632         * config/s390/s390-opts.h (enum processor_type):
8633         Remove enum values.
8634         * config/s390/s390.c
8635         (processor_table): Remove entries, add arch_level values.
8636         (s390_issue_rate): Remove cases.
8637         (s390_option_override): Adjust
8638         s390_option_override_internal() call.
8639         (s390_option_override_internal): Remove deprecation warning.
8640         (s390_valid_target_attribute_tree): Adjust
8641         s390_option_override_internal() call.
8642         * config/s390/s390.h (struct s390_processor):
8643         Share with s390-c.c, add arch_level field.
8644         * config/s390/s390.md:
8645         Remove occurrences in cpu attribute.
8646         * config/s390/s390.opt: Remove -march/-mtune support.
8647         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
8648         * doc/invoke.texi: Remove deprecation warning.
8650 2018-08-08  Luis Machado  <luis.machado@linaro.org>
8652         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
8653         global.
8654         (qdf24xx_tunings): Set vector cost structure to
8655         qdf24xx_vector_cost.
8657         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
8658         <register_sextend>: Set to 3.
8660 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
8662         PR target/86838
8663         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
8664         * config/aarch64/aarch64-simd.md
8665         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
8666         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
8667         (aarch64_frecpx<mode>): ...this new pattern.
8668         * config/aarch64/aarch64-simd-builtins.def: Remove comment
8669         about aarch64_frecp<FRECP:frecp_suffix><mode>.
8671 2018-08-07  Martin Liska  <mliska@suse.cz>
8673         PR middle-end/83023
8674         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
8675         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
8676         * predict.def (PRED_MALLOC_NONNULL): New predictor.
8677         * doc/extend.texi: Document that malloc attribute adds
8678         hit to compiler.
8680 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
8682         PR target/86785
8683         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8684         Define to speculation_safe_value_not_needed.
8686 2018-08-06  Jeff Law  <law@redhat.com>
8688         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
8689         the vr_values instance to cprop_into_stmt.
8690         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
8691         (cprop_operand): Also query EVRP to determine if OP is a constant.
8693 2018-08-06  Nathan Sidwell  <nathan@acm.org>
8695         * diagnostic.c (diagnostic_report_current_module): Reroll
8696         included-at loop.  Translate text.
8698 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
8700         * function-tests.c (selftest::test_expansion_to_rtl): Call
8701         free_after_compilation.
8703 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
8705         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
8707 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
8709         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
8710         loops with memory block operations from getting unrolled.
8712 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
8714         PR target/86807
8715         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8716         Define to speculation_safe_value_not_needed.
8718 2018-08-06  Jeff Law  <law@redhat.com>
8720         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
8721         assert.
8723 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
8725         PR target/86662
8726         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
8727         with all enabled __intN types.
8729         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
8731 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
8733         * alias.c (record_set): Check for clobber high.
8734         * cfgexpand.c (expand_gimple_stmt): Likewise.
8735         * combine-stack-adj.c (single_set_for_csa): Likewise.
8736         * combine.c (find_single_use_1): Likewise.
8737         (set_nonzero_bits_and_sign_copies): Likewise.
8738         (get_combine_src_dest): Likewise.
8739         (is_parallel_of_n_reg_sets): Likewise.
8740         (try_combine): Likewise.
8741         (record_dead_and_set_regs_1): Likewise.
8742         (reg_dead_at_p_1): Likewise.
8743         (reg_dead_at_p): Likewise.
8744         * dce.c (deletable_insn_p): Likewise.
8745         (mark_nonreg_stores_1): Likewise.
8746         (mark_nonreg_stores_2): Likewise.
8747         * df-scan.c (df_find_hard_reg_defs): Likewise.
8748         (df_uses_record): Likewise.
8749         (df_get_call_refs): Likewise.
8750         * dwarf2out.c (mem_loc_descriptor): Likewise.
8751         * haifa-sched.c (haifa_classify_rtx): Likewise.
8752         * ira-build.c (create_insn_allocnos): Likewise.
8753         * ira-costs.c (scan_one_insn): Likewise.
8754         * ira.c (equiv_init_movable_p): Likewise.
8755         (rtx_moveable_p): Likewise.
8756         (interesting_dest_for_shprep): Likewise.
8757         * jump.c (mark_jump_label_1): Likewise.
8758         * postreload-gcse.c (record_opr_changes): Likewise.
8759         * postreload.c (reload_cse_simplify): Likewise.
8760         (struct reg_use): Add source expr.
8761         (reload_combine): Check for clobber high.
8762         (reload_combine_note_use): Likewise.
8763         (reload_cse_move2add): Likewise.
8764         (move2add_note_store): Likewise.
8765         * print-rtl.c (print_pattern): Likewise.
8766         * recog.c (decode_asm_operands): Likewise.
8767         (store_data_bypass_p): Likewise.
8768         (if_test_bypass_p): Likewise.
8769         * regcprop.c (kill_clobbered_value): Likewise.
8770         (kill_set_value): Likewise.
8771         * reginfo.c (reg_scan_mark_refs): Likewise.
8772         * reload1.c (maybe_fix_stack_asms): Likewise.
8773         (eliminate_regs_1): Likewise.
8774         (elimination_effects): Likewise.
8775         (mark_not_eliminable): Likewise.
8776         (scan_paradoxical_subregs): Likewise.
8777         (forget_old_reloads_1): Likewise.
8778         * reorg.c (find_end_label): Likewise.
8779         (try_merge_delay_insns): Likewise.
8780         (redundant_insn): Likewise.
8781         (own_thread_p): Likewise.
8782         (fill_simple_delay_slots): Likewise.
8783         (fill_slots_from_thread): Likewise.
8784         (dbr_schedule): Likewise.
8785         * resource.c (update_live_status): Likewise.
8786         (mark_referenced_resources): Likewise.
8787         (mark_set_resources): Likewise.
8788         * rtl.c (copy_rtx): Likewise.
8789         * rtlanal.c (reg_referenced_p): Likewise.
8790         (single_set_2): Likewise.
8791         (noop_move_p): Likewise.
8792         (note_stores): Likewise.
8793         * sched-deps.c (sched_analyze_reg): Likewise.
8794         (sched_analyze_insn): Likewise.
8796 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
8798         * cse.c (invalidate_reg): New function extracted from...
8799         (invalidate): ...here.
8800         (canonicalize_insn): Check for clobber high.
8801         (invalidate_from_clobbers): invalidate clobber highs.
8802         (invalidate_from_sets_and_clobbers): Likewise.
8803         (count_reg_usage): Check for clobber high.
8804         (insn_live_p): Likewise.
8805         * cselib.c (cselib_expand_value_rtx_1):Likewise.
8806         (cselib_invalidate_regno): Check for clobber in setter.
8807         (cselib_invalidate_rtx): Pass through setter.
8808         (cselib_invalidate_rtx_note_stores):
8809         (cselib_process_insn): Check for clobber high.
8810         * cselib.h (cselib_invalidate_rtx): Add operand.
8812 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
8814         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
8815         (mark_not_eliminable): Likewise.
8816         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
8817         * lra-lives.c (process_bb_lives): Check for clobber high.
8818         * lra.c (new_insn_reg): Remember clobber highs.
8819         (collect_non_operand_hard_regs): Check for clobber high.
8820         (lra_set_insn_recog_data): Likewise.
8821         (add_regs_to_insn_regno_info): Likewise.
8822         (lra_update_insn_regno_info): Likewise.
8824 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
8826         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
8827         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
8829 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
8831         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
8832         (copy_insn_1): Likewise.
8833         (gen_hard_reg_clobber_high): New gen function.
8834         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
8835         * genemit.c (gen_exp): Likewise.
8836         (gen_emit_seq): Pass through info.
8837         (gen_insn): Check for CLOBBER_HIGH.
8838         (gen_expand): Pass through info.
8839         (gen_split): Likewise.
8840         (output_add_clobbers): Likewise.
8841         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
8842         (remove_clobbers): Likewise.
8843         * rtl.h (gen_hard_reg_clobber_high): New declaration.
8845 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
8847         * doc/rtl.texi (clobber_high): Add.
8848         (parallel): Add in clobber high
8849         * rtl.c (rtl_check_failed_code3): Add function.
8850         * rtl.def (CLOBBER_HIGH): Add expression.
8851         * rtl.h (RTL_CHECKC3): Add macro.
8852         (rtl_check_failed_code3): Add declaration.
8853         (XC3EXP): Add macro.
8855 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
8857         PR target/86386
8858         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
8859         cfun->machine->max_used_stack_alignment if needed.
8861 2018-08-04  Martin Sebor  <msebor@redhat.com>
8863         PR tree-optimization/86571
8864         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
8865         NaN output to 4.
8867 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
8869         PR target/86799
8870         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8871         Define.
8873 2018-08-03  Jeff Law  <law@redhat.com>
8875         PR target/86795
8876         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8877         Define to speculation_safe_value_not_needed.
8879 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
8881         * doc/gcov.texi (-x): Remove duplicate "to".
8882         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
8883         (-Wif-not-aligned): Remove duplicate "is".
8884         (-flto): Remove duplicate "the".
8885         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
8886         duplicate "v5.00.b".
8887         (MSP430 Options): Remove duplicate "and" from the description
8888         of "-mgprel-sec=regexp".
8889         (x86 Options): Remove duplicate copies of "vmldLog102" and
8890         vmlsLog104 from description of "-mveclibabi=type".
8892 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
8894         * internal-fn.h (first_commutative_argument): Declare.
8895         * internal-fn.c (first_commutative_argument): New function.
8896         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
8897         restrictions for pattern statements.  Use first_commutative_argument
8898         to look for commutative operands in calls to internal functions.
8900 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
8902         * Makefile.in (wide-int-range.o): New.
8903         * tree-vrp.c: Move all the wide_int_* functions to...
8904         * wide-int-range.cc: ...here.
8905         * tree-vrp.h: Move all the wide_int_* prototypes to...
8906         * wide-int-range.h: ...here.
8908 2018-08-03  Tom de Vries  <tdevries@suse.de>
8910         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
8911         UI_NONE.
8912         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
8913         * except.c (output_function_exception_table): Do early exit if
8914         targetm_common.except_unwind_info (&global_options) == UI_NONE.
8916 2018-08-03  Martin Liska  <mliska@suse.cz>
8918         * predict.c (dump_prediction): Change to 2 digits
8919         in fraction part.
8921 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
8923         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
8924         neon_dup_q to...
8925         (falkor_am_1_gtov_gtov): ... a new insn reservation.
8927 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
8929         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
8930         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
8931         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
8932         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
8933         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
8935 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
8937         * diagnostic-show-locus.c (diagnostic_show_locus): Use
8938         pp_take_prefix when saving the existing prefix.
8939         * diagnostic.c (diagnostic_append_note): Likewise.
8940         * langhooks.c (lhd_print_error_function): Likewise.
8941         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
8942         param's type.  Free the existing prefix.
8943         (pp_take_prefix): New function.
8944         (pretty_printer::pretty_printer): Drop the prefix parameter.
8945         Rename the length parameter to match the comment.
8946         (pretty_printer::~pretty_printer): Free the prefix.
8947         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
8948         parameter.
8949         (struct pretty_printer): Drop the "const" from "prefix" field's
8950         type and clarify memory management.
8951         (pp_set_prefix): Drop the "const" from the 2nd param.
8952         (pp_take_prefix): New decl.
8954 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
8956         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
8957         for word_mode_ok here instead of passing as argument.
8958         (expand_block_compare): Change select_block_compare_mode() call.
8959         (expand_strncmp_gpr_sequence): New function.
8960         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
8962 2018-08-02  Jeff Law  <law@redhat.com>
8964         PR target/86790
8965         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8966         Define to speculation_safe_value_not_needed.
8968         PR target/86784
8969         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8970         Define to speculation_safe_value_not_needed.
8972 2018-08-02  Tom de Vries  <tdevries@suse.de>
8974         PR target/86660
8975         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
8976         function.  Return UI_TARGET unconditionally.
8977         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
8978         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
8980 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
8982         * genemit.c (print_overload_test): Fix typo.
8984 2018-08-02  Richard Biener  <rguenther@suse.de>
8986         PR tree-optimization/86816
8987         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
8988         which checks for value availability before querying it.
8989         (gvn_uses_equal): Use it.
8990         (same_succ_hash): Likewise.
8991         (gimple_equal_p): Likewise.
8993 2018-08-02  Nick Clifton  <nickc@redhat.com>
8995         PR target/86813
8996         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
8997         Define to speculation_safe_value_not_needed.
8999         PR target/86810
9000         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9001         Define to speculation_safe_value_not_needed.
9003         PR target/86810
9004         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9005         Define to speculation_safe_value_not_needed.
9007         PR target/86803
9008         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9009         Define to speculation_safe_value_not_needed.
9011         PR target/86797
9012         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9013         Define to speculation_safe_value_not_needed.
9015         PR target/86791
9016         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9017         Define to speculation_safe_value_not_needed.
9019         PR target/86789
9020         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9021         Define to speculation_safe_value_not_needed.
9023         PR target/86787
9024         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9025         Define to speculation_safe_value_not_needed.
9027         PR target/86782
9028         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
9029         speculation_safe_value_not_needed.
9031         PR target/86781
9032         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
9033         to speculation_safe_value_not_needed.
9035 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
9037         * doc/md.texi: Expand the documentation of instruction names
9038         to mention port-local uses.  Document '@' in pattern names.
9039         * read-md.h (overloaded_instance, overloaded_name): New structs.
9040         (mapping): Declare.
9041         (md_reader::handle_overloaded_name): New member function.
9042         (md_reader::get_overloads): Likewise.
9043         (md_reader::m_first_overload): New member variable.
9044         (md_reader::m_next_overload_ptr): Likewise.
9045         (md_reader::m_overloads_htab): Likewise.
9046         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
9047         m_next_overload_ptr and m_overloads_htab.
9048         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
9049         (get_mode_token, get_code_token, get_int_token): New functions.
9050         (map_attr_string): Add an optional argument that passes back
9051         the associated iterator.
9052         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
9053         (md_reader::handle_overloaded_name, add_overload_instance): New
9054         functions.
9055         (apply_iterators): Handle '@' names.  Report an error if '@'
9056         is used without iterators.
9057         (initialize_iterators): Initialize the new iterator_group fields.
9058         * genopinit.c (handle_overloaded_code_for)
9059         (handle_overloaded_gen): New functions.
9060         (main): Use them to print declarations of maybe_code_for_* and
9061         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
9062         * genemit.c (print_overload_arguments, print_overload_test)
9063         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
9064         (main): Use it to print definitions of maybe_code_for_* and
9065         maybe_gen_* functions.
9066         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
9067         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
9068         instead of explicit mode checks.
9069         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
9070         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
9071         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
9072         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
9073         (aarch64_expand_compare_and_swap): Likewise
9074         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
9075         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
9076         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
9077         (aarch64_constant_pool_reload_icode): Delete.
9078         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
9079         instead of aarch64_constant_pool_reload_icode.  Use
9080         code_for_aarch64_reload_mov instead of explicit mode checks.
9081         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
9082         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
9083         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
9084         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
9085         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
9086         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
9087         (aarch64_atomic_load_op_code): Delete.
9088         (aarch64_emit_atomic_load_op): Likewise.
9089         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
9090         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
9091         instead of aarch64_emit_atomic_load_op.
9092         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
9093         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
9094         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
9095         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
9096         character before the pattern name.
9097         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
9098         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
9099         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
9100         (aarch64_frecps<mode>): Likewise.
9101         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
9102         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
9103         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
9104         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
9105         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
9107 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
9109         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
9110         Allow HFmode constants if TARGET_FP_F16INST.
9112 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
9114         PR target/86014
9115         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
9116         No longer check last store for clobber of address register.
9118 2018-08-02  Martin Liska  <mliska@suse.cz>
9120         PR gcov-profile/86817
9121         * gcov.c (process_all_functions): New function.
9122         (main): Call it.
9123         (process_file): Move functions processing to
9124         process_all_functions.
9126 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
9128         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
9129         "const" to the "gimple *" and "rtx_insn *" parameters.
9130         * dumpfile.h (dump_user_location_t::dump_user_location_t):
9131         Likewise.
9132         (dump_location_t::dump_location_t): Likewise.
9134 2018-08-01  Martin Sebor  <msebor@redhat.com>
9136         PR tree-optimization/86650
9137         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
9138         rather than a "gcall *".  Directly pass the data of interest
9139         to percent_K_format, rather than building a temporary CALL_EXPR
9140         to hold it.
9141         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
9142         (gimple_fold_builtin_strncat): Adjust.
9143         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
9144         gcall* argument with gimple*.
9145         * gimple-ssa-warn-restrict.c (check_call): Same.
9146         (wrestrict_dom_walker::before_dom_children): Same.
9147         (builtin_access::builtin_access): Same.
9148         (check_bounds_or_overlap): Same
9149         (maybe_diag_overlap): Same.
9150         (maybe_diag_offset_bounds): Same.
9151         * tree-diagnostic.c (default_tree_printer): Move usage of
9152         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
9153         to this callsite.
9154         * tree-pretty-print.c (percent_K_format): Add argument.
9155         * tree-pretty-print.h: Add argument.
9156         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
9157         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
9158         (maybe_diag_stxncpy_trunc): Same.
9159         (handle_builtin_stxncpy): Same.
9160         (handle_builtin_strcat): Same.
9162 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9164         * match.pd: Optimise pointer range checks.
9166 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9168         PR tree-optimization/86758
9169         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
9170         to remove pattern statements.
9172 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9174         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
9175         result of dfs_enumerate_from when constructing stmt_vec_infos,
9176         instead of additionally calling get_loop_body.
9178 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9180         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
9181         parameter.
9182         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
9183         When creating an iv, assert that the step is not known to be zero.
9184         (vect_setup_realignment): Update call accordingly.
9185         * tree-vect-stmts.c (vectorizable_store): Likewise.
9186         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
9188 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9190         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
9191         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
9192         (vectorizable_reduction): Likewise.
9193         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9194         (vect_detect_hybrid_slp_stmts): Likewise.
9195         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
9197 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
9199         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
9200         (wide_int_set_zero_nonzero_bits): ...this.
9201         (zero_nonzero_bits_from_vr): Rename to...
9202         (vrp_set_zero_nonzero_bits): ...this.
9203         (extract_range_from_multiplicative_op_1): Abstract wide int
9204         code...
9205         (wide_int_range_multiplicative_op): ...here.
9206         (extract_range_from_binary_expr_1): Extract wide int binary
9207         operations into their own functions.
9208         (wide_int_range_lshift): New.
9209         (wide_int_range_can_optimize_bit_op): New.
9210         (wide_int_range_shift_undefined_p): New.
9211         (wide_int_range_bit_xor): New.
9212         (wide_int_range_bit_ior): New.
9213         (wide_int_range_bit_and): New.
9214         (wide_int_range_trunc_mod): New.
9215         (extract_range_into_wide_ints): New.
9216         (vrp_shift_undefined_p): New.
9217         (extract_range_from_multiplicative_op): New.
9218         (vrp_can_optimize_bit_op): New.
9219         * tree-vrp.h (value_range::dump): New.
9220         (wide_int_range_multiplicative_op): New.
9221         (wide_int_range_lshift):New.
9222         (wide_int_range_shift_undefined_p): New.
9223         (wide_int_range_bit_xor): New.
9224         (wide_int_range_bit_ior): New.
9225         (wide_int_range_bit_and): New.
9226         (wide_int_range_trunc_mod): New.
9227         (zero_nonzero_bits_from_bounds): Rename to...
9228         (wide_int_set_zero_nonzero_bits): ...this.
9229         (zero_nonzero_bits_from_vr): Rename to...
9230         (vrp_set_zero_nonzero_bits): ...this.
9231         (range_easy_mask_min_max): Rename to...
9232         (wide_int_range_can_optimize_bit_op): this.
9233         * vr-values.c (simplify_bit_ops_using_ranges): Rename
9234         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
9236 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9238         * tree-vectorizer.h (vect_orig_stmt): New function.
9239         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
9240         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
9241         (vect_create_epilog_for_reduction): Likewise.
9242         (vectorizable_live_operation): Likewise.
9243         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
9244         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
9245         * tree-vect-stmts.c (vectorizable_call): Likewise.
9246         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
9248 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9250         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
9251         argument.
9252         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
9253         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
9254         (vect_transform_loop): Likewise.
9255         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
9257 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9259         * tree-vectorizer.h (vect_schedule_slp): Return void.
9260         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
9261         (vect_schedule_slp): Likewise.
9263 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9265         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
9266         argument.
9267         (vect_transform_loop): Update calls accordingly.  Schedule SLP
9268         instances before the main loop, if any exist.
9270 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
9272         PR tree-optimization/86749
9273         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
9274         If the lhs is used in a COND_EXPR, check that it is being used
9275         as the "then" or "else" value.
9277 2018-08-01  Tom de Vries  <tdevries@suse.de>
9279         PR target/86800
9280         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
9281         speculation_safe_value_not_needed.
9283 2018-08-01  Richard Biener  <rguenther@suse.de>
9285         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
9286         as base and offset.
9288 2018-08-01  Martin Liska  <mliska@suse.cz>
9290         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
9291         format how successful transformation is dumped.
9292         (gimple_mod_pow2_value_transform): Likewise.
9293         (gimple_mod_subtract_transform): Likewise.
9294         (gimple_stringops_transform): Likewise.
9296 2018-08-01  Martin Liska  <mliska@suse.cz>
9298         PR value-prof/35543
9299         * value-prof.c (interesting_stringop_to_profile_p):
9300         Simplify the code and add BUILT_IN_MEMMOVE.
9301         (gimple_stringops_transform): Likewise.
9303 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
9305         * config/aarch64/aarch64-simd.md
9306         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
9307         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
9308         use GPI iterator instead of SI mode.
9310 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
9312         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
9313         rs6000_speculation_barrier.
9314         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
9315         new barrier pattern name.
9317 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
9319         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
9320         (speculation_barrier): New insn.
9322 2018-08-01  Richard Biener  <rguenther@suse.de>
9324         PR bootstrap/86724
9325         * graphite.h: Include isl/id.h and isl/space.h to allow build
9326         with ISL 0.20.
9328 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
9330         PR target/86651
9331         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
9332         mode for COFF targets.
9333         * defaults.h (TARGET_COFF): Define.
9334         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
9335         TARGET_COFF): Define.
9336         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
9337         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
9338         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
9340 2018-07-31  Alexandre Oliva <oliva@adacore.com>
9342         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
9343         * gimple-streamer-out.c (output_bb): Save it.
9344         * lto-streamer-in.c (input_struct_function_base): Restore
9345         instance discriminator if available.  Create map on demand.
9346         * lto-streamer-out.c (output_struct_function_base): Save it if
9347         available.
9348         * final.c (decl_to_instance_map): Document LTO strategy.
9350 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
9351             Olivier Hainque  <hainque@adacore.com>
9353         * debug.h (decl_to_instance_map_t): New type.
9354         (decl_to_instance_map): Declare.
9355         (maybe_create_decl_to_instance_map): New inline function.
9356         * final.c (bb_discriminator, last_bb_discriminator): New statics,
9357         to track basic block discriminators.
9358         (final_start_function_1): Initialize them.
9359         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
9360         bb_discriminator.
9361         (decl_to_instance_map): New variable.
9362         (map_decl_to_instance, maybe_set_discriminator): New functions.
9363         (notice_source_line): Set discriminator.
9365 2018-07-31  Ian Lance Taylor  <iant@golang.org>
9367         * targhooks.c (default_have_speculation_safe_value): Add
9368         ATTRIBUTE_UNUSED.
9370 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
9372         * dump-context.h: Include "pretty-print.h".
9373         (dump_context::refresh_dumps_are_enabled): New decl.
9374         (dump_context::emit_item): New decl.
9375         (class dump_context): Add fields "m_test_pp" and
9376         "m_test_pp_flags".
9377         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
9378         (temp_dump_context::get_dumped_text): New decl.
9379         (class temp_dump_context): Add field "m_pp".
9380         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
9381         (dump_context::refresh_dumps_are_enabled): ...and add a test for
9382         m_test_pp.
9383         (set_dump_file): Update for above change.
9384         (set_alt_dump_file): Likewise.
9385         (dump_loc): New overload, taking a pretty_printer *.
9386         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
9387         to any test pretty-printer.
9388         (make_item_for_dump_gimple_stmt): New function, adapted from
9389         optinfo::add_gimple_stmt.
9390         (dump_context::dump_gimple_stmt): Call it, and use the result,
9391         eliminating the direct usage of dump_file and alt_dump_file in
9392         favor of indirectly using them via emit_item.
9393         (make_item_for_dump_gimple_expr): New function, adapted from
9394         optinfo::add_gimple_expr.
9395         (dump_context::dump_gimple_expr): Call it, and use the result,
9396         eliminating the direct usage of dump_file and alt_dump_file in
9397         favor of indirectly using them via emit_item.
9398         (make_item_for_dump_generic_expr): New function, adapted from
9399         optinfo::add_tree.
9400         (dump_context::dump_generic_expr): Call it, and use the result,
9401         eliminating the direct usage of dump_file and alt_dump_file in
9402         favor of indirectly using them via emit_item.
9403         (make_item_for_dump_printf_va): New function, adapted from
9404         optinfo::add_printf_va.
9405         (make_item_for_dump_printf): New function.
9406         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
9407         and use the result, eliminating the direct usage of dump_file and
9408         alt_dump_file in favor of indirectly using them via emit_item.
9409         (make_item_for_dump_dec): New function.
9410         (dump_context::dump_dec): Call it, and use the result,
9411         eliminating the direct usage of dump_file and alt_dump_file in
9412         favor of indirectly using them via emit_item.
9413         (make_item_for_dump_symtab_node): New function, adapted from
9414         optinfo::add_symtab_node.
9415         (dump_context::dump_symtab_node): Call it, and use the result,
9416         eliminating the direct usage of dump_file and alt_dump_file in
9417         favor of indirectly using them via emit_item.
9418         (dump_context::begin_scope): Reimplement, avoiding direct usage
9419         of dump_file and alt_dump_file in favor of indirectly using them
9420         via emit_item.
9421         (dump_context::emit_item): New member function.
9422         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
9423         Set up test pretty-printer on the underlying context.  Call
9424         refresh_dumps_are_enabled.
9425         (temp_dump_context::~temp_dump_context): Call
9426         refresh_dumps_are_enabled.
9427         (temp_dump_context::get_dumped_text): New member function.
9428         (selftest::verify_dumped_text): New function.
9429         (ASSERT_DUMPED_TEXT_EQ): New macro.
9430         (selftest::test_capture_of_dump_calls): Run all tests twice, with
9431         and then without optinfo enabled.  Add uses of
9432         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
9433         * dumpfile.h: Update comment for the dump_* API.
9434         * optinfo-emit-json.cc
9435         (selftest::test_building_json_from_dump_calls): Update for new
9436         param for temp_dump_context ctor.
9437         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
9438         and "m_owned" field.
9439         (optinfo_item::~optinfo_item): Likewise.
9440         (optinfo::add_item): New member function.
9441         (optinfo::emit): Update comment.
9442         (optinfo::add_string): Delete.
9443         (optinfo::add_printf): Delete.
9444         (optinfo::add_printf_va): Delete.
9445         (optinfo::add_gimple_stmt): Delete.
9446         (optinfo::add_gimple_expr): Delete.
9447         (optinfo::add_tree): Delete.
9448         (optinfo::add_symtab_node): Delete.
9449         (optinfo::add_dec): Delete.
9450         * optinfo.h (class dump_context): New forward decl.
9451         (optinfo::add_item): New decl.
9452         (optinfo::add_string): Delete.
9453         (optinfo::add_printf): Delete.
9454         (optinfo::add_printf_va): Delete.
9455         (optinfo::add_gimple_stmt): Delete.
9456         (optinfo::add_gimple_expr): Delete.
9457         (optinfo::add_tree): Delete.
9458         (optinfo::add_symtab_node): Delete.
9459         (optinfo::add_dec): Delete.
9460         (optinfo::add_poly_int): Delete.
9461         (optinfo_item::optinfo_item): Remove "owned" param.
9462         (class optinfo_item): Remove field "m_owned".
9464 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
9466         PR middle-end/86705
9467         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
9468         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
9469         requested variable alignment.
9470         (expand_one_ssa_partition): Likewise.
9471         (expand_one_var): Likewise.
9473 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9475         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
9476         to speculation_safe_value_not_needed.
9478 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9480         * targhooks.h (speculation_safe_value_not_needed): New prototype.
9481         * targhooks.c (speculation_safe_value_not_needed): New function.
9482         * target.def (have_speculation_safe_value): Update documentation.
9483         * doc/tm.texi: Regenerated.
9485 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9487         * config/aarch64/iterators.md (ALLI_TI): New iterator.
9488         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
9489         expand.
9490         (despeculate_copy<ALLI:mode>_insn): New insn.
9491         (despeculate_copyti_insn): New insn.
9492         (despeculate_simple<ALLI:mode>): New insn
9493         (despeculate_simpleti): New insn.
9494         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
9495         function.
9496         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
9497         aarch64_speculation_safe_value.
9498         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
9500 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9502         * config/aarch64/aarch64-speculation.cc: New file.
9503         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
9504         before pass_reorder_blocks.
9505         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
9506         prototype.
9507         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
9508         X14 and X15 when tracking speculation.
9509         * config/aarch64/aarch64.md (register name constants): Add
9510         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
9511         (unspec): Add UNSPEC_SPECULATION_TRACKER.
9512         (speculation_barrier): New insn attribute.
9513         (cmp<mode>): Allow SP in comparisons.
9514         (speculation_tracker): New insn.
9515         (speculation_barrier): Add speculation_barrier attribute.
9516         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
9517         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
9518         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
9520 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9522         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
9523         aarch64_track_speculation is true.
9524         (tb<optab><mode>1): Likewise.
9525         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
9526         generate CB[N]Z when tracking speculation.
9527         (aarch64_split_compare_and_swap): Likewise.
9528         (aarch64_split_atomic_op): Likewise.
9530 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9532         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
9534 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9536         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
9537         (speculation_barrier): New insn.
9539 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9541         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
9542         * config/arm/arm.md (speculation_barrier): New expand.
9543         (speculation_barrier_insn): New pattern.
9545 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
9547         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
9548         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
9549         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
9550         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
9551         list.
9552         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
9553         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
9554         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
9555         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
9556         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
9557         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
9558         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
9559         * builtins.c (expand_speculation_safe_value): New function.
9560         (expand_builtin): Call it.
9561         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
9562         * doc/extend.texi: Document __builtin_speculation_safe_value.
9563         * doc/md.texi: Document "speculation_barrier" pattern.
9564         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
9565         TARGET_HAVE_SPECULATION_SAFE_VALUE.
9566         * doc/tm.texi: Regenerated.
9567         * target.def (have_speculation_safe_value, speculation_safe_value): New
9568         hooks.
9569         * targhooks.c (default_have_speculation_safe_value): New function.
9570         (default_speculation_safe_value): New function.
9571         * targhooks.h (default_have_speculation_safe_value): Add prototype.
9572         (default_speculation_safe_value): Add prototype.
9574 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
9576         * dump-context.h (dump_context::dump_loc): New decl.
9577         * dumpfile.c (dump_context::dump_loc): New member function.
9578         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
9579         and dump_gimple_stmt.
9580         (dump_context::dump_gimple_expr_loc): Likewise, using
9581         dump_gimple_expr.
9582         (dump_context::dump_generic_expr_loc): Likewise, using
9583         dump_generic_expr.
9584         (dump_context::dump_printf_loc_va): Likewise, using
9585         dump_printf_va.
9586         (dump_context::begin_scope): Explicitly using the global function
9587         "dump_loc", rather than the member function.
9589 2018-07-31  Martin Sebor  <msebor@redhat.com>
9591         PR tree-optimization/86741
9592         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
9594 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
9596         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
9597         depend on whether prefetch instructions will be emitted or not.
9598         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
9599         will be emitted or not.
9600         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
9601         (TARGET_SETMEM_PFD): New macros.
9603 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9605         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
9606         (NULL_STMT_VEC_INFO): Delete.
9607         (stmt_vec_info::operator*): Likewise.
9608         (stmt_vec_info::operator gimple *): Likewise.
9609         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
9610         of NULL_STMT_VEC_INFO.
9611         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
9612         (vect_reassociating_reduction_p): Likewise.
9613         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
9614         (vectorizable_store): Likewise.
9615         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
9616         (vec_info::free_stmt_vec_infos): Likewise.
9618 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9620         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
9621         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
9623 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9625         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
9626         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
9627         (vec_info::free_stmt_vec_info): New private member functions.
9628         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
9629         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
9630         * tree-parloops.c (gather_scalar_reductions): Remove calls to
9631         set_stmt_vec_info_vec and free_stmt_vec_infos.
9632         * tree-vect-loop.c (_loop_vec_info): Remove call to
9633         set_stmt_vec_info_vec.
9634         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
9635         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
9636         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
9637         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
9638         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
9639         assignments in {vec_info::,}new_stmt_vec_info that are redundant
9640         with the clearing in the xcalloc.
9641         (stmt_vec_info_vec): Delete.
9642         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
9643         (vectorize_loops): Likewise.
9644         (vec_info::~vec_info): Remove argument from call to
9645         free_stmt_vec_infos.
9646         (vec_info::add_stmt): Remove vinfo argument from call to
9647         new_stmt_vec_info.
9649 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9651         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
9652         rather than a gimple stmt.
9653         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
9654         information for pattern statements when passed the original
9655         statement; instead wait to be passed the pattern statement itself.
9656         Don't call set_vinfo_for_stmt here.
9657         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
9658         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
9659         stmt_vec_infos here.
9660         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
9661         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
9662         stmt_vec_infos entry.
9664 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9666         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
9667         * tree-vectorizer.c (vec_info::replace_stmt): New function.
9668         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
9669         * tree-vect-stmts.c (vectorizable_call): Likewise.
9670         (vectorizable_simd_clone_call): Likewise.
9672 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9674         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
9675         * tree-vectorizer.c (vec_info::remove_stmt): New function.
9676         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
9677         * tree-vect-loop.c (vect_transform_loop): Likewise.
9678         * tree-vect-slp.c (vect_schedule_slp): Likewise.
9679         * tree-vect-stmts.c (vect_remove_stores): Likewise.
9681 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9683         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
9684         (vect_dr_stmt): Delete.
9685         * tree-vectorizer.c (vec_info::lookup_dr): New function.
9686         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
9687         of DR_VECT_AUX.
9688         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9689         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
9690         (vect_verify_datarefs_alignment, vect_peeling_supportable)
9691         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
9692         (vect_analyze_data_refs): Likewise.
9693         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
9694         argument.
9695         (vect_find_same_alignment_drs): Likewise.
9696         (vect_slp_analyze_node_dependences): Update calls accordingly.
9697         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
9698         instead of DR_VECT_AUX.
9699         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
9700         of a vector data references.  Use vec_info::lookup_dr instead of
9701         DR_VECT_AUX.
9702         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
9703         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
9704         instead of DR_VECT_AUX.
9706 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9708         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
9709         dr_vec_info.
9710         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
9711         accordingly.
9712         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
9713         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
9714         (vect_gen_prolog_loop_niters): Likewise.
9716 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9718         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
9719         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
9720         (vect_known_alignment_in_bytes, vect_dr_behavior)
9721         (vect_get_scalar_dr_size): Take references as dr_vec_infos
9722         instead of data_references.  Update calls to other routines for
9723         which the same change has been made.
9724         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
9725         dr_vec_infos instead of stmt_vec_infos.
9726         (vect_analyze_data_ref_dependence): Update call accordingly.
9727         (vect_slp_analyze_data_ref_dependence)
9728         (vect_record_base_alignments): Use DR_VECT_AUX.
9729         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
9730         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
9731         (vector_alignment_reachable_p, vect_get_data_access_cost)
9732         (vect_peeling_supportable, vect_analyze_group_access_1)
9733         (vect_analyze_group_access, vect_analyze_data_ref_access)
9734         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
9735         (vect_compile_time_alias, vect_small_gap_p)
9736         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
9737         (vect_supportable_dr_alignment): Take references as dr_vec_infos
9738         instead of data_references.  Update calls to other routines for
9739         which the same change has been made.
9740         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
9741         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
9742         (vect_slp_analyze_and_verify_node_alignment)
9743         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
9744         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
9745         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
9746         above changes.
9747         (_vect_peel_info::dr): Replace with...
9748         (_vect_peel_info::dr_info): ...this new field.
9749         (vect_peeling_hash_get_most_frequent)
9750         (vect_peeling_hash_choose_best_peeling): Update accordingly.
9751         (vect_peeling_hash_get_lowest_cost):
9752         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
9753         routines for which the same change has been made.
9754         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
9755         data_reference.
9756         * tree-vect-loop-manip.c (get_misalign_in_elems)
9757         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
9758         above changes.
9759         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
9760         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
9761         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
9762         (get_group_load_store_type, get_negative_load_store_type)
9763         (vect_get_data_ptr_increment, vectorizable_store)
9764         (vectorizable_load): Likewise.
9765         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
9766         Update calls to other routines for which the same change has been made.
9768 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9770         * tree-vectorizer.h (vec_info::move_dr): New member function.
9771         (dataref_aux): Rename to...
9772         (dr_vec_info): ...this and add "dr" and "stmt" fields.
9773         (_stmt_vec_info::dr_aux): Update accordingly.
9774         (_stmt_vec_info::data_ref_info): Delete.
9775         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
9776         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
9777         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
9778         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
9779         of data_ref.
9780         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
9781         (STMT_VINFO_DR_INFO): New macro.
9782         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
9783         (set_dr_misalignment): Update after rename of dataref_aux.
9784         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
9785         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
9786         initialization of STMT_VINFO_DATA_REF.
9787         * tree-vectorizer.c (vec_info::move_dr): New function.
9788         * tree-vect-patterns.c (vect_recog_bool_pattern)
9789         (vect_recog_mask_conversion_pattern)
9790         (vect_recog_gather_scatter_pattern): Use it.
9791         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
9792         the "dr" and "stmt" fields of dr_vec_info instead of
9793         STMT_VINFO_DATA_REF.
9795 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9797         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
9798         (is_pattern_stmt_p): Use it.
9799         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
9800         on pattern statements.
9802 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9804         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
9805         original stmt as a stmt_vec_info rather than a gimple stmt.
9806         (vect_pattern_recog_1): Take the statement directly as a
9807         stmt_vec_info, rather than via a gimple_stmt_iterator.
9808         Update call to vect_mark_pattern_stmts.
9809         (vect_pattern_recog): Update calls accordingly.
9811 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9813         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
9814         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
9815         a vect_def_type for the first argument.
9816         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
9817         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
9818         operand if it isn't defined by a vectorized statement.
9819         (vect_build_gather_load_calls): Remove the mask_dt argument and
9820         update calls to vect_get_vec_def_for_stmt_copy.
9821         (vectorizable_bswap): Likewise the dt argument.
9822         (vectorizable_call): Update calls to vectorizable_bswap and
9823         vect_get_vec_def_for_stmt_copy.
9824         (vectorizable_simd_clone_call, vectorizable_assignment)
9825         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
9826         (vectorizable_comparison): Update calls to
9827         vect_get_vec_def_for_stmt_copy.
9828         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
9829         vect_is_simple_use.
9830         (vect_get_loop_based_defs): Remove dt argument and update call
9831         to vect_get_vec_def_for_stmt_copy.
9832         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
9833         and vect_get_vec_def_for_stmt_copy.
9834         (vectorizable_load): Update calls to vect_build_gather_load_calls
9835         and vect_get_vec_def_for_stmt_copy.
9836         * tree-vect-loop.c (vect_create_epilog_for_reduction)
9837         (vectorizable_reduction, vectorizable_live_operation): Update calls
9838         to vect_get_vec_def_for_stmt_copy.
9840 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9842         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
9843         and gimple stmt arguments with a stmt_vec_info.
9844         (vect_record_base_alignments): Update calls accordingly.
9845         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
9846         and gimple stmt arguments with a stmt_vec_info.
9847         (vect_build_slp_tree_1): Remove vinfo argument and update call
9848         to vect_record_max_nunits.
9849         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
9850         and vect_record_max_nunits.
9852 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9854         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
9855         file and take a stmt_vec_info instead of a gimple stmt.
9856         (supportable_widening_operation, vect_finish_replace_stmt)
9857         (vect_finish_stmt_generation, vect_get_store_rhs)
9858         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
9859         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
9860         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
9861         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
9862         (vect_create_data_ref_ptr, bump_vector_ptr)
9863         (vect_permute_store_chain, vect_setup_realignment)
9864         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
9865         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
9866         (vectorizable_reduction, vectorizable_induction)
9867         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
9868         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
9869         than gimple stmts as arguments.
9870         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
9871         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
9872         (can_group_stmts_p, vect_check_gather_scatter)
9873         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
9874         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
9875         (vect_permute_load_chain, vect_shift_permute_load_chain)
9876         (vect_transform_grouped_load)
9877         (vect_record_grouped_load_vectors): Likewise.
9878         * tree-vect-loop.c (vect_fixup_reduc_chain)
9879         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
9880         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
9881         (vectorizable_reduction, vectorizable_induction)
9882         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
9883         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
9884         (vect_get_load_store_mask): Likewise.
9885         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
9886         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
9887         * tree-vect-stmts.c (vect_mark_relevant)
9888         (is_simple_and_all_uses_invariant)
9889         (exist_non_indexing_operands_for_use_p, process_use)
9890         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
9891         (vect_get_vec_def_for_operand, vect_get_vec_defs)
9892         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
9893         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
9894         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
9895         (get_negative_load_store_type, get_load_store_type)
9896         (vect_check_load_store_mask, vect_check_store_rhs)
9897         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
9898         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
9899         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
9900         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9901         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
9902         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
9903         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
9904         (supportable_widening_operation): Likewise.
9906 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9908         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
9909         a stmt_vec_info instead of a gcall.
9910         (vect_check_gather_scatter): Update call accordingly.
9911         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
9912         of a gphi.
9913         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
9914         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
9915         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
9916         instead of a gimple stmt.
9917         (vect_transform_loop): Update calls accordingly.
9918         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
9919         stmt_vec_infos instead of gimple stmts.
9920         (vect_analyze_slp_instance): Update use accordingly.
9921         * tree-vect-stmts.c (read_vector_array, write_vector_array)
9922         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
9923         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
9924         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
9925         (vect_gen_widened_results_half, vect_get_loop_based_defs)
9926         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
9927         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
9928         down to subroutines.
9930 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9932         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
9933         of the worklist from a vector of gimple stmts to a vector of
9934         stmt_vec_infos.
9935         * tree-vect-stmts.c (vect_mark_relevant, process_use)
9936         (vect_mark_stmts_to_be_vectorized): Likewise
9938 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9940         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
9941         statement before passing it to vect_analyze_stmt.
9942         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
9943         the chain of phi vector definitions.  Track the exit phi via its
9944         stmt_vec_info.
9945         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
9946         STMT_VINFO_REDUC_DEF.
9947         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
9948         stmt_vec_infos to handle the statement chains.
9949         (vect_get_slp_defs): Record the first statement in the node
9950         using a stmt_vec_info.
9951         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
9952         statements here and pass their stmt_vec_info down to subroutines.
9953         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
9954         down to vect_finish_stmt_generation.
9955         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
9956         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
9957         stmt_vec_infos to subroutines.
9958         (vect_remove_stores): Use stmt_vec_infos to handle the statement
9959         chains.
9961 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9963         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
9964         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
9965         (vect_permute_store_chain, vect_setup_realignment)
9966         (vect_permute_load_chain, vect_shift_permute_load_chain)
9967         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
9968         stmts internally, and when passing values to other vectorizer routines.
9969         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
9970         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
9971         (vect_analyze_loop_operations, get_initial_def_for_reduction)
9972         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
9973         (vectorizable_reduction, vectorizable_induction)
9974         (vectorizable_live_operation, vect_transform_loop_stmt)
9975         (vect_transform_loop): Likewise.
9976         * tree-vect-patterns.c (vect_reassociating_reduction_p)
9977         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
9978         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
9979         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9980         (vect_slp_analyze_node_operations_1): Likewise.
9981         * tree-vect-stmts.c (vect_mark_relevant, process_use)
9982         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
9983         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
9984         (vect_finish_stmt_generation_1, get_group_load_store_type)
9985         (get_load_store_type, vect_build_gather_load_calls)
9986         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
9987         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
9988         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
9989         (vectorizable_store, vectorizable_load, vectorizable_condition)
9990         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
9991         (supportable_widening_operation): Likewise.
9992         (vect_get_vector_types_for_stmt): Likewise.
9993         * tree-vectorizer.h (vect_dr_behavior): Likewise.
9995 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
9997         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9998         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
9999         (vect_permute_store_chain, vect_permute_load_chain)
10000         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
10001         repeated stmt_vec_info lookups.
10002         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
10003         (vect_update_ivs_after_vectorizer): Likewise.
10004         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
10005         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
10006         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
10007         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10008         (vect_bb_slp_scalar_cost): Likewise.
10009         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
10011 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10013         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
10014         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
10015         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
10016         to get gassigns and gcalls, rather than operating on generc gimple
10017         stmts.
10018         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
10019         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
10020         (vectorizable_load, vect_analyze_stmt): Likewise.
10021         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
10023 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10025         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
10026         return stmt_vec_infos rather than gimple stmts.  Do not accept
10027         null arguments.
10028         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
10029         of a gimple stmt.
10030         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10031         Update use of get_later_stmt.
10032         (vect_get_constant_vectors): Update call accordingly.
10033         (vect_schedule_slp_instance): Likewise
10034         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
10035         (vect_slp_analyze_instance_dependence): Likewise.
10036         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
10038 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10040         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
10041         (stmt_info_for_cost::stmt_info): ...this new field.
10042         (add_stmt_costs): Update accordingly.
10043         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
10044         (vect_get_known_peeling_cost): Likewise.
10045         (vect_estimate_min_profitable_iters): Likewise.
10046         * tree-vect-stmts.c (record_stmt_cost): Likewise.
10048 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10050         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
10051         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10052         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10053         accordingly.
10054         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
10056 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10058         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
10059         a gimple stmt to a stmt_vec_info.
10060         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
10062 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10064         * tree-vectorizer.h (vec_info::grouped_stores): Change from
10065         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10066         (_loop_vec_info::reduction_chains): Likewise.
10067         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
10068         accordingly.
10069         * tree-vect-slp.c (vect_analyze_slp): Likewise.
10071 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10073         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
10074         a gimple stmt to a stmt_vec_info.
10075         (_stmt_vec_info::next_element): Likewise.
10076         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
10077         (vect_slp_analyze_and_verify_node_alignment)
10078         (vect_analyze_group_access_1, vect_analyze_group_access)
10079         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
10080         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
10081         (vect_supportable_dr_alignment): Update accordingly.
10082         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
10083         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
10084         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
10085         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
10086         * tree-vect-slp.c (vect_build_slp_tree_1)
10087         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
10088         (vect_split_slp_store_group, vect_analyze_slp_instance)
10089         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
10090         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
10091         (get_group_load_store_type, get_load_store_type)
10092         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
10093         (vect_transform_stmt, vect_remove_stores): Likewise.
10095 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10097         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
10098         than a gimple stmt.
10099         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10100         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
10101         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
10102         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
10103         (vector_alignment_reachable_p, vect_get_data_access_cost)
10104         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
10105         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
10106         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
10107         (vect_analyze_group_access_1, vect_analyze_group_access)
10108         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
10109         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
10110         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
10111         result of vect_dr_stmt and use the stmt_vec_info instead of
10112         the associated gimple stmt.
10113         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10114         (vect_gen_prolog_loop_niters): Likewise.
10115         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10117 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10119         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
10120         to a vec<stmt_vec_info>.
10121         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
10122         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
10123         vec<stmt_vec_info>.
10124         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
10125         to a vec<stmt_vec_info>.
10126         (bst_traits::value_type, bst_traits::value_type): Likewise.
10127         (bst_traits::hash): Update accordingly.
10128         (vect_get_and_check_slp_defs): Change the stmts parameter from
10129         a vec<gimple *> to a vec<stmt_vec_info>.
10130         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
10131         (vect_build_slp_tree): Likewise.
10132         (vect_build_slp_tree_2): Likewise.  Update uses of
10133         SLP_TREE_SCALAR_STMTS.
10134         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
10135         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
10136         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
10137         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
10138         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
10139         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
10140         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
10141         (vect_get_constant_vectors, vect_get_slp_defs)
10142         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
10143         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
10144         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
10145         instead of gimple stmts.
10146         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
10147         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
10148         (vect_slp_analyze_instance_dependence): Update uses of
10149         SLP_TREE_SCALAR_STMTS.
10150         (vect_slp_analyze_and_verify_node_alignment): Likewise.
10151         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
10152         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
10153         (get_initial_defs_for_reduction): Likewise.
10154         (vect_create_epilog_for_reduction): Likewise.
10155         (vectorize_fold_left_reduction): Likewise.
10156         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
10157         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
10158         (can_vectorize_live_stmts): Likewise.
10160 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10162         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
10163         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10164         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
10165         than gimple stmts.
10166         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
10167         of a gimple stmt.
10168         (gather_scalar_reductions): Update after above interface changes.
10169         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
10170         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
10171         than gimple stmts.
10172         (vect_force_simple_reduction): Likewise.
10173         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
10174         LOOP_VINFO_REDUCTIONS.
10175         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10177 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10179         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
10180         a gimple stmt to a stmt_vec_info.
10181         * tree-vect-loop.c (vect_active_double_reduction_p)
10182         (vect_force_simple_reduction, vectorizable_reduction): Update
10183         accordingly.
10185 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10187         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
10188         vec<gimple *> to a vec<stmt_vec_info>.
10189         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
10190         the reduction_phis argument from a vec<gimple *> to a
10191         vec<stmt_vec_info>.
10192         (vectorizable_reduction): Likewise the phis local variable that
10193         is passed to vect_create_epilog_for_reduction.  Update for new type
10194         of SLP_TREE_VEC_STMTS.
10195         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
10196         (vectorizable_live_operation): Likewise.
10197         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
10198         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
10200 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10202         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
10203         a gimple stmt to a stmt_vec_info.
10204         (vectorizable_condition, vectorizable_live_operation)
10205         (vectorizable_reduction, vectorizable_induction): Pass back the
10206         vectorized statement as a stmt_vec_info.
10207         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
10208         use of STMT_VINFO_VEC_STMT.
10209         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
10210         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
10211         as stmt_vec_infos rather than gimple stmts.
10212         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
10213         to a stmt_vec_info.
10214         (vectorizable_live_operation): Likewise.
10215         (vectorizable_reduction, vectorizable_induction): Likewise,
10216         updating use of STMT_VINFO_VEC_STMT.
10217         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
10218         of STMT_VINFO_VEC_STMT.
10219         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
10220         (vectorizable_simd_clone_call, vectorizable_conversion)
10221         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10222         (vectorizable_store, vectorizable_load, vectorizable_condition)
10223         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
10224         from a gimple stmt to a stmt_vec_info.
10225         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
10226         pointer to a stmt_vec_info to the vectorizable_* routines.
10228 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10230         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
10231         a gimple stmt to a stmt_vec_info.
10232         (is_pattern_stmt_p): Update accordingly.
10233         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
10234         (vect_record_grouped_load_vectors): Likewise.
10235         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
10236         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
10237         (vect_model_reduction_cost): Likewise.
10238         (vect_create_epilog_for_reduction): Likewise.
10239         (vectorizable_reduction, vectorizable_induction): Likewise.
10240         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10241         Return the stmt_vec_info for the pattern statement.
10242         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
10243         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
10244         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
10245         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
10246         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
10247         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
10248         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
10249         (free_stmt_vec_info, vect_is_simple_use): Likewise.
10251 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10253         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
10254         (vect_finish_stmt_generation): Likewise.
10255         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
10256         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
10257         (vect_build_gather_load_calls): Use the return value of the above
10258         functions instead of a separate call to vinfo_for_stmt.  Use narrow
10259         scopes for the input gimple stmt and wider scopes for the associated
10260         stmt_vec_info.  Use vec_info::lookup_def when setting these
10261         stmt_vec_infos from an SSA_NAME definition.
10262         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10263         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10264         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10265         (vectorizable_store, vectorizable_load, vectorizable_condition)
10266         (vectorizable_comparison): Likewise.
10267         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
10268         (vectorizable_reduction): Likewise.
10270 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10272         * tree-vectorizer.h (vect_is_simple_use): Add an optional
10273         stmt_vec_info * parameter before the optional gimple **.
10274         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10275         (process_use, vect_get_vec_def_for_operand_1): Update callers.
10276         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
10277         * tree-vect-loop.c (vectorizable_reduction): Likewise.
10278         (vectorizable_live_operation): Likewise.
10279         * tree-vect-patterns.c (type_conversion_p): Likewise.
10280         (vect_look_through_possible_promotion): Likewise.
10281         (vect_recog_rotate_pattern): Likewise.
10282         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
10284 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10286         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
10287         a typedef to a wrapper class.
10288         (NULL_STMT_VEC_INFO): New macro.
10289         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
10290         (stmt_vec_info::operator*): New function.
10291         (stmt_vec_info::operator gimple *): Likewise.
10292         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
10293         (add_stmt_costs): Likewise.
10294         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
10295         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
10296         (vect_get_known_peeling_cost): Likewise.
10297         (vect_estimate_min_profitable_iters): Likewise.
10298         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10299         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
10300         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
10301         (vectorizable_store, free_stmt_vec_infos): Likewise.
10302         (new_stmt_vec_info): Change return type of xcalloc to
10303         _stmt_vec_info *.
10305 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10307         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
10308         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
10309         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
10310         a single_imm_use-based sequence.
10311         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
10313 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10315         * tree-vectorizer.h (vec_info::lookup_def): Declare.
10316         * tree-vectorizer.c (vec_info::lookup_def): New function.
10317         * tree-vect-patterns.c (vect_get_internal_def): Use it.
10318         (vect_widened_op_tree): Likewise.
10319         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10320         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
10321         (vectorizable_reduction): Likewise.
10322         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
10323         of a gimple *.
10324         (vect_is_slp_reduction): Update calls accordingly.  Use
10325         vec_info::lookup_def.
10326         (vect_is_simple_reduction): Likewise
10327         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
10329 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10331         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
10332         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
10333         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
10334         of vinfo_for_stmt.
10335         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
10336         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
10337         (vect_update_vf_for_slp, vect_analyze_loop_operations)
10338         (vect_is_slp_reduction, vectorizable_induction)
10339         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
10340         * tree-vect-patterns.c (vect_init_pattern_stmt):
10341         (vect_determine_min_output_precision_1, vect_determine_precisions)
10342         (vect_pattern_recog): Likewise.
10343         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
10344         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
10345         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
10346         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
10347         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
10348         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
10349         info field from a loop to a loop_vec_info.
10351 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10353         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
10354         (vec_info::add_stmt): Declare.
10355         * tree-vectorizer.c (vec_info::add_stmt): New function.
10356         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
10357         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
10358         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
10359         (vectorizable_induction): Likewise.
10360         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
10361         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
10362         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
10363         (vectorizable_load): Likewise.
10364         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10365         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
10366         (vect_recog_gather_scatter_pattern): Likewise.
10367         (append_pattern_def_seq): Likewise.  Remove a check that is
10368         performed by add_stmt itself.
10370 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10372         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
10373         which make_ssa_name was called with new_stmt before new_stmt
10374         had been created.
10376 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10378         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
10379         split out from...
10380         (vect_is_slp_reduction): ...here...
10381         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
10382         that are already known to be false.
10384 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10386         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
10387         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
10388         STMT_VINFO_NUM_SLP_USES when it's true.
10389         (vect_free_slp_instance): Add a final_p parameter and pass it to
10390         vect_free_slp_tree.
10391         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
10392         (vect_analyze_slp_instance): Likewise.
10393         (vect_slp_analyze_operations): Likewise.
10394         (vect_slp_analyze_bb_1): Likewise.
10395         * tree-vectorizer.c (vec_info): Likewise.
10396         * tree-vect-loop.c (vect_transform_loop): Likewise.
10398 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10400         * tree-vect-loop.c (vectorizable_reduction): Assert that the
10401         function is not called for second and subsequent members of
10402         a reduction group.
10404 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10406         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
10407         cases for nested loops from here to ...
10408         (vect_create_epilog_for_reduction): ...here.  Only call
10409         vect_is_simple_use for inner-loop reductions.
10411 2018-07-31  Martin Liska  <mliska@suse.cz>
10413         PR gcov-profile/85338
10414         PR gcov-profile/85350
10415         PR gcov-profile/85372
10416         * profile.c (struct location_triplet): New.
10417         (struct location_triplet_hash): Likewise.
10418         (output_location): Do not output a BB that
10419         is already recorded for a line.
10420         (branch_prob): Use streamed_locations.
10422 2018-07-31  Martin Liska  <mliska@suse.cz>
10424         PR gcov-profile/85370
10425         * coverage.c (coverage_begin_function): Do not mark target
10426         clones as artificial functions.
10428 2018-07-31  Martin Liska  <mliska@suse.cz>
10430         PR gcov-profile/83813
10431         PR gcov-profile/84758
10432         PR gcov-profile/85217
10433         PR gcov-profile/85332
10434         * profile.c (branch_prob): Do not record GOTO expressions
10435         for GIMPLE statements which locations are already streamed.
10437 2018-07-31  Olivier Hainque  <hainque@adacore.com>
10439         * gcc.c (handle_spec_function): Accept a soft_matched_part
10440         argument, as do_spec_1.  Pass it down to ...
10441         (eval_spec_function): Accept a soft_matched_part argument,
10442         and pass it down to ...
10443         (do_spec_2): Accept a soft_matched_part argument, and pass
10444         it down to do_spec_1.
10445         (do_spec_1): Pass soft_matched_part to handle_spec_function.
10446         (handle_braces): Update call to handle_spec_function.
10447         (driver::set_up_specs): Update calls to do_spec_2.
10448         (compare_debug_dump_opt_spec_function): Likewise.
10449         (compare_debug_self_opt_spec_function): Likewise.
10451 2018-07-31  Olivier Hainque  <hainque@adacore.com>
10453         * common.opt (nolibc): New option.
10454         * doc/invoke.texi (Link Options): Document it.
10455         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
10456         * config/alpha/linux.h: Likewise.
10457         * config/arc/elf.h: Likewise.
10458         * config/arm/uclinux-elf.h: Likewise.
10459         * config/arm/unknown-elf.h: Likewise.
10460         * config/avr/avrlibc.h: Likewise.
10461         * config/bfin/bfin.h: Likewise.
10462         * config/bfin/linux.h: Likewise.
10463         * config/bfin/uclinux.h: Likewise.
10464         * config/darwin.h: Likewise.
10465         * config/darwin10.h: Likewise.
10466         * config/darwin12.h: Likewise.
10467         * config/gnu-user.h: Likewise.
10468         * config/lm32/uclinux-elf.h: Likewise.
10469         * config/pa/pa-hpux11.h: Likewise.
10470         * config/pa/pa64-hpux.h: Likewise.
10471         * config/sparc/sparc.h: Likewise.
10473 2018-07-31  Olivier Hainque  <hainque@adacore.com>
10475         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
10476         undefined variables.
10478 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
10480         PR target/86640
10481         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
10482         instead of GEN_INT.
10484 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10486         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
10487         terminated string literal.
10489 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
10491         PR rtl-optimization/85160
10492         * combine.c (is_just_move): New function.
10493         (try_combine): Allow combining two instructions into two if neither of
10494         the original instructions was a move.
10496 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
10498         PR target/86673
10499         * doc/extend.texi (Global Register Variables): Discourage use of type
10500         qualifiers.
10501         (Local Register Variables): Likewise.
10503 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
10505         PR tree-optimization/86506
10506         * hwint.h (ceil_log2): Resync with hwint.c implementation.
10508 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
10510         PR target/86547
10511         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
10512         hard_regno, make sure no insn between `from` and `to` clobbers it.
10514 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
10515             Tom de Vries  <tdevries@suse.de>
10517         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
10518         (PTX_DEFAULT_RUNTIME_DIM): ... this.
10519         (nvptx_goacc_validate_dims): Set default worker and gang dims to
10520         PTX_DEFAULT_RUNTIME_DIM.
10521         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
10523 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
10525         * config/pa/pa.c (pa_output_addr_vec): Align address table.
10526         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
10527         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
10529 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
10531         * config/rs6000/constraints.md (wG constraint): Delete, no longer
10532         used.
10533         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
10534         predicate to reflect toc fusion has been deleted.
10535         (toc_fusion_mem_raw): Delete, no longer used.
10536         (toc_fusion_mem_wrapped): Likewise.
10537         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
10538         fusion mask bit.
10539         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
10540         Delete, no longer used.
10541         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
10542         meant to be used for toc fusion.
10543         (rs6000_debug_print_mode): Delete toc fusion debugging.
10544         (rs6000_debug_reg_global): Likewise.
10545         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
10546         fusion and secondary reload support that were never used.
10547         (rs6000_option_override_internal): Delete TOC fusion, that was only
10548         partially defined, and it did not work unless you also used the
10549         -mcmodel= switch.
10550         (rs6000_legitimate_address_p): Delete TOC fusion support.
10551         (rs6000_opt_masks): Likewise.
10552         (fusion_wrap_memory_address): Delete function, no longer used.
10553         (fusion_split_address); Delete TOC fusion support.
10554         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
10555         longer used with toc fusion being deleted.
10556         (TARGET_TOC_FUSION_FP): Likewise.
10557         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
10558         UNSPEC.
10559         (toc fusion spliter): Delete TOC fusion support.
10560         (toc_fusionload_<mode>): Likewise.
10561         (toc_fusionload_di): Likewise.
10562         (fusion_gpr_load_<mode>): Delete generator function, this insn no
10563         longer needs to be named.  Rename predicate to delete TOC fusion.
10564         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
10565         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
10566         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
10567         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
10568         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
10570 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10572         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
10573         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
10574         __int128 in built-in function prototypes.
10575         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
10576         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
10578 2018-07-27  Martin Sebor  <msebor@redhat.com>
10580         PR tree-optimization/86696
10581         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
10582         types, including enums.
10583         (handle_char_store): Be prepared for the above function to fail.
10585 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
10587         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
10588         when optimization level is lower than 2 or optimize for size.
10590 2018-07-26  Martin Sebor  <msebor@redhat.com>
10592         PR tree-optimization/86043
10593         PR tree-optimization/86042
10594         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
10595         (get_string_cst_length): Rename...
10596         (get_min_string_length): ...to this.  Add argument.
10597         (handle_char_store): Extend to handle multi-character stores by
10598         MEM_REF.
10599         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
10600         * tree.h (initializer_zerop): Add argument.
10602 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
10604         PR middle-end/86660
10605         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
10606         declare target to variables if they have always,{to,from,tofrom} map
10607         kinds.
10609 2018-07-26  Martin Liska  <mliska@suse.cz>
10611         PR lto/86548
10612         * lto-wrapper.c: Add linker_output as prefix
10613         for ltrans_output_file.
10615 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
10617         PR rtl-optimization/85805
10618         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
10619         value for hard registers if that was written in the same mode.
10621 2018-07-26  Martin Liska  <mliska@suse.cz>
10623         PR gcov-profile/86536
10624         * gcov.c (format_gcov): Use printf format %.*f directly
10625         and do not handle special values.
10627 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
10629         * common/config/arc/arc-common.c (arc_option_optimization_table):
10630         Update default optimizations for size.
10632 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
10634         * config/arc/arc.md (movsf_insn): Add short instruction selection.
10635         * config/arc/constraints.md (CfZ): New constraint.
10636         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
10637         (subsf3_fpu): Likewise.
10638         (cmpsf_fpu): Likewise.
10639         (cmpsf_fpu_uneq): Likewise.
10641 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
10643         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
10644         accesses as well.
10645         (arc_is_uncached_mem_p): uncached applies to both the variable and
10646         the pointer.
10648 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
10650         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
10651         register names.
10653 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
10655         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
10656         field "m_scopes" from vec to auto_vec.
10658 2018-07-25  Martin Liska  <mliska@suse.cz>
10660         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
10661         return type.
10663 2018-07-25  Richard Biener  <rguenther@suse.de>
10665         PR debug/86654
10666         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
10667         special wrt context_die late.
10668         (gen_subprogram_die): Re-use DIEs in local scope.
10670 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
10672         PR tree-optimization/86644
10673         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
10675 2018-07-25  Martin Liska  <mliska@suse.cz>
10677         PR middle-end/86645
10678         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
10679         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
10681 2018-07-25  Martin Liska  <mliska@suse.cz>
10683         PR sanitizer/79635
10684         * params.def: Explain ASan abbreviation and provide
10685         a documentation link.
10687 2018-07-24  Martin Sebor  <msebor@redhat.com>
10689         PR tree-optimization/86622
10690         PR tree-optimization/86532
10691         * builtins.h (string_length): Declare.
10692         * builtins.c (c_strlen): Correct handling of non-constant offsets.
10693         (check_access): Be prepared for non-constant length ranges.
10694         (string_length): Make extern.
10695         * expr.c (string_constant): Only handle the minor non-constant
10696         array index.  Use string_constant to compute the length of
10697         a generic string constant.
10699 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
10701         PR tree-optimization/86618
10702         * tree-vect-stmts.c (vectorizable_call): Don't take the address
10703         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
10705 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
10707         PR tree-optimization/86636
10708         * json.cc (json::object::set): Fix comment.  Add assertions.
10709         (json::array::append): Move here from json.h.  Add comment and an
10710         assertion.
10711         (json::string::string): Likewise.
10712         * json.h (json::array::append): Move to json.cc.
10713         (json::string::string): Likewise.
10714         * optinfo-emit-json.cc
10715         (optrecord_json_writer::impl_location_to_json): Assert that we
10716         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
10717         wrapper around it.  Expand the location once, rather than three
10718         times.
10719         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
10720         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
10721         wrappers.
10722         (optrecord_json_writer::optinfo_to_json): Likewise, in four
10723         places.  Fix some overlong lines.
10725 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
10727         * config/aarch64/aarch64-simd.md
10728         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
10729         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
10730         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
10731         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
10732         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
10733         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
10734         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
10735         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
10736         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
10738 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
10740         PR middle-end/86627
10741         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
10742         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
10743         and abs_d == d, do the power of two handling if profitable.
10745 2018-07-24  Richard Biener  <rguenther@suse.de>
10747         * match.pd: Add BIT_FIELD_REF canonicalizations.
10749 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10751         PR c/86617
10752         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
10754 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10756         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
10757         terminated STRING_CST object.
10759 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10761         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
10763 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
10765         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
10766         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
10767         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
10768         the elements into a register.
10769         (rs6000_split_v4si_init_di_reg): Delete.
10770         (rs6000_split_v4si_init): Delete.
10771         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
10772         (vsx_init_v4si): Rewrite as a define_expand.
10774 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
10776         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
10777         zero_extend argument from memory): New.
10779 2018-07-22  Martin Sebor  <msebor@redhat.com>
10781         PR bootstrap/86621
10782         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
10783         diagnosing calls with unknown arguments unless -Walloca-larger-than
10784         is restricted to less than PTRDIFF_MAX bytes.
10786 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
10788         * doc/gcov.texi (Invoking Gcov): Editorial changes.
10790 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
10792         * pretty-print.c (text_info::set_location): Remove redundant
10793         "line_table" parameter from call to rich_location::set_range.
10795 2018-07-20  Martin Sebor  <msebor@redhat.com>
10797         PR middle-end/82063
10798         * builtins.c (expand_builtin_alloca): Adjust.
10799         * calls.c (alloc_max_size): Simplify.
10800         * cgraphunit.c (cgraph_node::expand): Adjust.
10801         * common.opt (larger_than_size, warn_frame_larger_than): Remove
10802         variables.
10803         (frame_larger_than_size): Same.
10804         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
10805         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
10806         Initialize.
10807         * doc/invoke.texi (GCC Command Options): Document option arguments.
10808         Explain byte-size arguments and suffixes.
10809         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
10810         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
10811         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
10812         * doc/options.texi (UInteger): Expand.
10813         (Host_Wide_Int, ByteSize): Document new properties.
10814         * final.c (final_start_function_1): Include sizes in an error message.
10815         * function.c (frame_offset_overflow): Same.
10816         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
10817         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
10818         Diagnose unbounded alloca calls only for limits of less than
10819         PTRDIFF_MAX.
10820         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
10821         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
10822         for alloca(0).
10823         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
10824         only for limits of less than PTRDIFF_MAX.
10825         * langhooks-def.h (lhd_handle_option): Change function argument
10826         to HOST_WIDE_INT.
10827         * langhooks.c (lhd_handle_option): Same.
10828         * langhooks.h (handle_option): Same.
10829         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
10830         ByteSize flags.
10831         (var_type, var_type_struct): Same.
10832         (var_set): Handle ByteSize flag.
10833         * optc-gen.awk: Add comments to output to ease debugging.  Make
10834         use of HOST_WIDE_INT where appropriate.
10835         * opts-gen-save.awk:  Use %lx to format unsigned long.
10836         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
10837         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
10838         arguments.  Parse bytes-size suffixes.
10839         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
10840         (enum_value_to_arg): Same.
10841         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
10842         (handle_option): Adjust.
10843         (generate_option): Change function argument to HOST_WIDE_INT.
10844         (cmdline_handle_error): Adjust.
10845         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
10846         (set_option): Change function argument to HOST_WIDE_INT.
10847         (option_enabled): Handle cl_host_wide_int.
10848         (get_option_state): Handle CLVC_SIZE.
10849         (control_warning_option): Same.
10850         * opts.c (common_handle_option): Change function argument to
10851         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
10852         OPT_Wvla_larger_than_.
10853         * opts.h (enum cl_var_type): Add an enumerator.
10854         * stor-layout.c (layout_decl): Print a more meaningful warning.
10855         * toplev.c (output_stack_usage): Adjust.
10857 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
10859         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
10860         call to inline_expand_builtin_string_cmp.
10861         (expand_builtin_strcmp): Likewise.
10862         (expand_builtin_strncmp): Likewise.
10863         (inline_string_cmp): Delete the last parameter, change char_type_node
10864         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
10865         two operands.
10866         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
10867         the inlining expansion on target where the type of the call has same or
10868         narrower precision than unsigned char.
10870 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
10872         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
10873         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
10874         * common.opt (fsave-optimization-record): New option.
10875         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
10876         * doc/invoke.texi (-fsave-optimization-record): New option.
10877         * dumpfile.c: Include "optinfo-emit-json.h".
10878         (struct kv_pair): Move to coretypes.h.
10879         (optgroup_options): Make non-static.
10880         (dump_context::end_scope): Call
10881         optimization_records_maybe_pop_dump_scope.
10882         * dumpfile.h (optgroup_options): New decl.
10883         * json.cc: New file.
10884         * json.h: New file.
10885         * optinfo-emit-json.cc: New file.
10886         * optinfo-emit-json.h: New file.
10887         * optinfo.cc: Include "optinfo-emit-json.h".
10888         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
10889         (optinfo_enabled_p): Check optimization_records_enabled_p.
10890         (optinfo_wants_inlining_info_p): Likewise.
10891         * optinfo.h: Update comment.
10892         * profile-count.c (profile_quality_as_string): New function.
10893         * profile-count.h (profile_quality_as_string): New decl.
10894         (profile_count::quality): New accessor.
10895         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
10896         and optinfo_emit_json_cc_tests.
10897         * selftest.h (selftest::json_cc_tests): New decl.
10898         (selftest::optinfo_emit_json_cc_tests): New decl.
10899         * toplev.c: Include "optinfo-emit-json.h".
10900         (compile_file): Call optimization_records_finish.
10901         (do_compile): Call optimization_records_start.
10902         * tree-ssa-live.c: Include optinfo.h.
10903         (remove_unused_scope_block_p): Retain inlining information if
10904         optinfo_wants_inlining_info_p returns true.
10906 2018-07-20  Richard Biener  <rguenther@suse.de>
10908         PR debug/86585
10909         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
10910         to cover -flto-partition=none.
10912 2018-07-20  Martin Liska  <mliska@suse.cz>
10914         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
10915         (get_decl_source_range): Remove unused function.
10917 2018-07-20  Richard Biener  <rguenther@suse.de>
10919         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
10920         (struct vn_phi_s): Likewise.
10921         (struct vn_reference_s): Likewise.
10922         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
10923         for searching the slot of an entry known to be in the hash itself.
10924         (vn_phi_hasher::equal): Likewise.
10925         (vn_reference_hasher::equal): Likewise.
10926         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
10927         globals.
10928         (optimistic_info, current_info): Remove, keeping only valid_info.
10929         (vn_reference_lookup_1): Remove fallback lookup.
10930         (vn_reference_lookup_2): Likewise.
10931         (vn_nary_op_lookup_1): Likewise.
10932         (vn_phi_lookup): Likewise.
10933         (vn_nary_build_or_lookup_1): Make sure to not chain the built
10934         hash element.
10935         (vn_reference_insert): Adjust, chain the inserted hash element
10936         at last_inserted_ref.
10937         (vn_reference_insert_pieces): Likewise.
10938         (visit_reference_op_call): Likewise.
10939         (vn_nary_op_insert_into): Chain the inserted hash element at
10940         last_inserted_nary.
10941         (vn_nary_op_insert_pieces): Adjust.
10942         (vn_nary_op_insert): Likewise.
10943         (vn_nary_op_insert_stmt): Likewise.
10944         (vn_phi_insert): Adjust, chain the inserted hash element at
10945         last_inserted_phi.
10946         (process_scc): Remove clearing and copying the optimistic
10947         table.  Instead remove elements inserted during an optimistic
10948         iteration from the single table we maintain.
10949         (init_scc_vn): Adjust.
10950         (free_scc_vn): Likewise.
10951         (sccvn_dom_walker::record_cond): Likewise.
10952         (sccvn_dom_walker::after_dom_children): Likewise.
10954 2018-07-19  Martin Sebor  <msebor@redhat.com>
10956         PR tree-optimization/84047
10957         PR tree-optimization/83776
10958         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
10959         (check_array_bounds): Call it.
10961 2018-07-19  Martin Sebor  <msebor@redhat.com>
10963         * align.h (align_flags): Use member initialization.
10965 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
10967         * Makefile.in (OBJS): Add optinfo.o.
10968         * coretypes.h (class symtab_node): New forward decl.
10969         (struct cgraph_node): New forward decl.
10970         (class varpool_node): New forward decl.
10971         * dump-context.h: New file.
10972         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
10973         "tree-pass.h".
10974         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
10975         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
10976         (set_alt_dump_file): Likewise.
10977         (dump_context::~dump_context): New dtor.
10978         (dump_gimple_stmt): Move implementation to...
10979         (dump_context::dump_gimple_stmt): ...this new member function.
10980         Add the stmt to any pending optinfo, creating one if need be.
10981         (dump_gimple_stmt_loc): Move implementation to...
10982         (dump_context::dump_gimple_stmt_loc): ...this new member function.
10983         Start a new optinfo and add the stmt to it.
10984         (dump_gimple_expr): Move implementation to...
10985         (dump_context::dump_gimple_expr): ...this new member function.
10986         Add the stmt to any pending optinfo, creating one if need be.
10987         (dump_gimple_expr_loc): Move implementation to...
10988         (dump_context::dump_gimple_expr_loc): ...this new member function.
10989         Start a new optinfo and add the stmt to it.
10990         (dump_generic_expr): Move implementation to...
10991         (dump_context::dump_generic_expr): ...this new member function.
10992         Add the tree to any pending optinfo, creating one if need be.
10993         (dump_generic_expr_loc): Move implementation to...
10994         (dump_context::dump_generic_expr_loc): ...this new member
10995         function.  Add the tree to any pending optinfo, creating one if
10996         need be.
10997         (dump_printf): Move implementation to...
10998         (dump_context::dump_printf_va): ...this new member function.  Add
10999         the text to any pending optinfo, creating one if need be.
11000         (dump_printf_loc): Move implementation to...
11001         (dump_context::dump_printf_loc_va): ...this new member function.
11002         Start a new optinfo and add the stmt to it.
11003         (dump_dec): Move implementation to...
11004         (dump_context::dump_dec): ...this new member function.  Add the
11005         value to any pending optinfo, creating one if need be.
11006         (dump_context::dump_symtab_node): New member function.
11007         (dump_context::get_scope_depth): New member function.
11008         (dump_context::begin_scope): New member function.
11009         (dump_context::end_scope): New member function.
11010         (dump_context::ensure_pending_optinfo): New member function.
11011         (dump_context::begin_next_optinfo): New member function.
11012         (dump_context::end_any_optinfo): New member function.
11013         (dump_context::s_current): New global.
11014         (dump_context::s_default): New global.
11015         (dump_scope_depth): Delete global.
11016         (dumpfile_ensure_any_optinfo_are_flushed): New function.
11017         (dump_symtab_node): New function.
11018         (get_dump_scope_depth): Reimplement in terms of dump_context.
11019         (dump_begin_scope): Likewise.
11020         (dump_end_scope): Likewise.
11021         (selftest::temp_dump_context::temp_dump_context): New ctor.
11022         (selftest::temp_dump_context::~temp_dump_context): New dtor.
11023         (selftest::verify_item): New function.
11024         (ASSERT_IS_TEXT): New macro.
11025         (ASSERT_IS_TREE): New macro.
11026         (ASSERT_IS_GIMPLE): New macro.
11027         (selftest::test_capture_of_dump_calls): New test.
11028         (selftest::dumpfile_c_tests): Call it.
11029         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
11030         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
11031         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
11032         descriptive comment.
11033         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
11034         (dump_node, dump_bb): Move these unrelated decls.
11035         (class dump_manager): Add leading comment.
11036         * optinfo.cc: New file.
11037         * optinfo.h: New file.
11039 2018-07-19  Michael Collison  <michael.collison@arm.com>
11040             Richard Henderson <rth@redhat.com>
11042         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
11043         (subti): Handle op1 zero.
11044         (subvti4, usub4ti4): New.
11045         (*sub<GPI>3_compare1_imm): New.
11046         (sub<GPI>3_carryinCV): New.
11047         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
11048         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
11050 2018-07-19  Michael Collison  <michael.collison@arm.com>
11051             Richard Henderson <rth@redhat.com>
11053         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
11054         (addti3): Create simpler code if low part is already known to be 0.
11055         (addvti4, uaddvti4): New.
11056         (*add<GPI>3_compareC_cconly_imm): New.
11057         (*add<GPI>3_compareC_cconly): New.
11058         (*add<GPI>3_compareC_imm): New.
11059         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
11060         handle constants within this pattern..
11061         (*add<GPI>3_compareV_cconly_imm): New.
11062         (*add<GPI>3_compareV_cconly): New.
11063         (*add<GPI>3_compareV_imm): New.
11064         (add<GPI>3_compareV): New.
11065         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
11066         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
11067         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
11068         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
11069         with 'comparison' operator.
11070         (*add<GPI>3_compareV_cconly_imm): Ditto.
11071         (*add<GPI>3_compareV_cconly): Ditto.
11072         (*add<GPI>3_compareV_imm): Ditto.
11073         (add<GPI>3_compareV): Ditto.
11074         (add<mode>3_carryinC): Ditto.
11075         (*add<mode>3_carryinC_zero): Ditto.
11076         (*add<mode>3_carryinC): Ditto.
11077         (add<mode>3_carryinV): Ditto.
11078         (*add<mode>3_carryinV_zero): Ditto.
11079         (*add<mode>3_carryinV): Ditto.
11081 2018-07-19  Michael Collison  <michael.collison@arm.com>
11082             Richard Henderson <rth@redhat.com>
11084         * config/aarch64/aarch64-modes.def (CC_V): New.
11085         * config/aarch64/aarch64-protos.h
11086         (aarch64_addti_scratch_regs): Declare
11087         (aarch64_subvti_scratch_regs): Declare.
11088         (aarch64_expand_subvti): Declare.
11089         (aarch64_gen_unlikely_cbranch): Declare
11090         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
11091         for signed overflow using CC_Vmode.
11092         (aarch64_get_condition_code_1): Handle CC_Vmode.
11093         (aarch64_gen_unlikely_cbranch): New function.
11094         (aarch64_addti_scratch_regs): New function.
11095         (aarch64_subvti_scratch_regs): New function.
11096         (aarch64_expand_subvti): New function.
11098 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11100         * config/aarch64/aarch64-option-extensions.def: New entry for profile
11101         extension.
11102         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
11103         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
11104         extension.
11106 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11108         PR target/83009
11109         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11110         address check not strict.
11112 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11114         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
11115         Umq with Umn.
11116         (store_pair_lanes<mode>): Likewise.
11117         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
11118         enum value 'ADDR_QUERY_LDP_STP_N'.
11119         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
11120         (aarch64_print_address_internal): Add declaration.
11121         (aarch64_print_ldpstp_address): Remove.
11122         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
11123         (aarch64_print_operand): Change printing of 'y'.
11124         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
11125         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
11126         'true' rather than '1'.
11127         * config/aarch64/constraints.md (Uml): Likewise.
11128         (Uml): Rename to Umn.
11129         (Umq): Remove.
11131 2018-07-19  Richard Biener  <rguenther@suse.de>
11133         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
11134         a trailing array.
11135         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
11136         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
11137         (vn_reference_hasher): Likewise.
11138         (struct vn_tables_s): Remove obstack and alloc-pool members.
11139         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
11140         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
11141         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
11142         (vn_reference_insert_pieces): Likewise.
11143         (alloc_vn_nary_op_noinit): Adjust.
11144         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
11145         (vn_phi_eq): Adjust.
11146         (shared_lookup_phiargs): Remove.
11147         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
11148         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
11149         (visit_reference_op_call): Likewise.
11150         (copy_nary, copy_phi, copy_reference): Remove.
11151         (process_scc): Rewind the obstack when iterating.  Do not
11152         copy the elements to valid_info but just move them from one
11153         hashtable to the other.
11154         (allocate_vn_table): Adjust.
11155         (free_vn_table): Likewise.
11156         (init_scc_vn): Likewise.
11157         (free_scc_vn): Likewise.
11159 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
11161         PR target/86560
11162         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
11163         indirect_return as function type attribute.
11164         (ix86_attribute_table): Change indirect_return to function
11165         type attribute.
11166         * doc/extend.texi: Update indirect_return attribute.
11168 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
11170         * wide-int.h (widest2_int): New.
11171         * gimple-fold.c (arith_overflowed_p): Use it.
11172         * tree.h (widest2_int_cst): New.
11173         * tree-vrp.c (wide_int_binop_overflow): Rename from
11174         vrp_int_const_binop.
11175         Rewrite to work on trees.
11176         (extract_range_from_multiplicative_op_1): Abstract code to...
11177         (wide_int_range_min_max): ...here.
11178         (wide_int_range_cross_product): ...and here.
11179         (extract_range_from_binary_expr_1): Abstract overflow code to...
11180         (wide_int_range_mult_wrapping): ...here.
11181         * tree-vrp.h (wide_int_range_cross_product): New.
11182         (wide_int_range_mult_wrapping): New.
11184 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
11185             Julia Koval  <julia.koval@intel.com>
11187         * config/i386/x86-tune-costs.h (skylake_memcpy,
11188         skylake_memset): Replace rep_prefix with unrolling for size 512.
11190 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
11192         PR middle-end/86544
11193         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
11194         comparision with EQ_EXPR in last stmt.
11196 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11198         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
11199         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
11200         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
11201         previously known as "PowerPC AltiVec Built-in Functions".  Move
11202         some material to new subsubsections "PowerPC AltiVec Built-in
11203         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
11204         ISA 2.07".
11205         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
11206         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
11207         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
11208         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
11210 2018-07-18  Richard Biener  <rguenther@suse.de>
11212         PR tree-optimization/86557
11213         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
11214         EXACT_DIV_EXPR.
11216 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
11218         * config/s390/s390.c (s390_function_profiler): Generate CFI.
11220 2018-07-17  Jeff Law  <law@redhat.com>
11222         * config/arm/arm.c (get_label_padding): Update for recent
11223         changes to label_to_alignment.
11225         PR tree-optimization/86010
11226         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
11228         * config/mips/mips.c (vr4130_align_insns): Update for recent
11229         changes to label_to_alignment.
11231         * config/frv/frv.c (frv_label_align): Update for recent changes
11232         to label_to_alignment.
11234         * config/nios2/nios2.c (nios2_label_align): Update for recent
11235         changes which dropped ALIGN_LABELS_LOG.
11237 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
11239         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
11240         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
11242 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
11244         * config/arc/arc.c (arc_label_align): Use align_labels instead of
11245         deprecated align_labels_log.
11247 2018-07-17  Richard Biener  <rguenther@suse.de>
11249         PR lto/86456
11250         * dwarf2out.c (init_sections_and_labels): Always generate
11251         a debug_line_str_section for early LTO debug.
11252         (dwarf2out_finish): Reset debug_line_str_hash output early.
11253         Bump counter for extra dwarf5 .debug_loc labels to not conflict
11254         with fat LTO part.
11255         (dwarf2out_early_finish): Output debug_line_str.
11257 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
11259         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
11260         index register on z196 or later.
11262 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
11264         * config/s390/s390.c (s390_default_align): Set default function
11265         alignment to 16.
11266         (s390_override_options_after_change): Call s390_default align.
11267         (s390_option_override_internal): Call s390_default_align.
11268         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
11270 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
11272         PR middle-end/86542
11273         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
11274         to _looptemp_ clauses, other than the first two.
11276 2018-07-17  Martin Liska  <mliska@suse.cz>
11278         * opts.c: Do not enable OPT_falign_* for -Os.
11280 2018-07-17  Martin Liska  <mliska@suse.cz>
11282         * align.h (MAX_CODE_ALIGN): New.
11283         (MAX_CODE_ALIGN_VALUE): New.
11284         * common/config/i386/i386-common.c (ix86_handle_option):
11285         (MAX_CODE_ALIGN): Moved to align.h.
11286         * final.c (MAX_CODE_ALIGN): Likewise.
11287         * opts.c (parse_and_check_align_values):
11288         (MAX_CODE_ALIGN): Likewise.
11289         (MAX_CODE_ALIGN_VALUE): Likewise.
11291 2018-07-17  Martin Liska  <mliska@suse.cz>
11293         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
11294         in order to fulfil coding style.
11295         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
11296         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11297         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
11298         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
11299         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
11300         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
11301         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
11302         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
11303         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
11305 2018-07-17  Martin Liska  <mliska@suse.cz>
11307         * align.h: New file.
11308         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
11309         directly.
11310         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
11311         align_flags of label_to_alignment.
11312         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
11313         align_flags class.
11314         * config/m68k/m68k.c: Do not use removed align_labels_value and
11315         align_loops_value.
11316         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
11317         (LOOP_ALIGN): Likewise.
11318         (LABEL_ALIGN): Likewise.
11319         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
11320         Remove not used macro.
11321         (rs6000_loop_align): Change return type to align_flags.
11322         (rs6000_loop_align_max_skip): Remove.
11323         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
11324         Change return type to align_flags.
11325         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
11326         Remove not used macro.
11327         (rs6000_loop_align):  Change return type to align_flags.
11328         (rs6000_loop_align_max_skip): Remove.
11329         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
11330         * config/rx/rx-protos.h (rx_align_for_label): Make it
11331         static function.
11332         * config/rx/rx.c (rx_align_for_label): Change return type
11333         to align_flags.
11334         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
11335         macro definitions.
11336         into align_flags class.
11337         (LABEL_ALIGN): Likewise.
11338         (LOOP_ALIGN): Likewise.
11339         * config/s390/s390.c (s390_label_align): Use align_flags
11340         class member.
11341         (s390_asm_output_function_label): Likewise.
11342         * config/sh/sh.c (sh_override_options_after_change):
11343         Use align_flags class directly without macros.
11344         (find_barrier): Likewise.
11345         (barrier_align): Likewise.
11346         (sh_loop_align): Likewise.
11347         * config/spu/spu.c (spu_option_override):
11348         Use align_flags_tuple::get_value instead of removed macros.
11349         (spu_sched_init): Likewise.
11350         * config/spu/spu.h (GTY): Likewise.
11351         * config/visium/visium.c (visium_option_override):
11352         Set "8" as default secondary alignment.
11353         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
11354         in order to guarantee secondary alignment of 8.
11355         * coretypes.h: Include align.h header file.
11356         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
11357         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
11358         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
11359         * doc/tm.texi.in: Likewise.
11360         * final.c (struct label_alignment): Remove not used structure.
11361         (LABEL_ALIGN): Change type to align_flags.
11362         (LOOP_ALIGN): Likewise.
11363         (JUMP_ALIGN): Likewise.
11364         (default_loop_align_max_skip): Remove.
11365         (default_label_align_max_skip): Likewise.
11366         (default_jump_align_max_skip): Likewise.
11367         (default_label_align_after_barrier_max_skip):
11368         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
11369         (LABEL_TO_MAX_SKIP): Remove.
11370         (label_to_alignment): Return align_flags type instead of integer.
11371         (label_to_max_skip): Remove.
11372         (align_fuzz): Use align_flags type.
11373         (compute_alignments): Use align_flags type and use align_flags::max
11374         to combine multiple alignments.
11375         (grow_label_align): Grow vec instead of C array.
11376         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
11377         (shorten_branches):  Use align_flags type and use align_flags::max
11378         to combine multiple alignments.
11379         (final_scan_insn_1): Remove usage of secondary alignment that comes
11380         from label alignment, but instead use proper secondary alignment
11381         which is computed in grow_label_align.
11382         * flags.h (struct align_flags_tuple): Move to align.h.
11383         (struct align_flags): Likewise.
11384         (state_align_loops): Rename to align_loops.
11385         (state_align_jumps): Rename to align_jumps.
11386         (state_align_labels): Rename to align_labels.
11387         (state_align_functions): Rename to align_functions.
11388         (align_loops_log): Remove.
11389         (align_jumps_log): Remove.
11390         (align_labels_log): Remove.
11391         (align_functions_log): Remove.
11392         (align_loops_max_skip): Remove.
11393         (align_jumps_max_skip): Remove.
11394         (align_labels_max_skip): Remove.
11395         (align_functions_max_skip): Remove.
11396         (align_loops_value): Remove.
11397         (align_jumps_value): Remove.
11398         (align_labels_value): Remove.
11399         (align_functions_value): Remove.
11400         * output.h (label_to_alignment): Change return type to align_flags.
11401         (label_to_max_skip): Remove.
11402         * target.def: Remove loop_align_max_skip, label_align_max_skip,
11403         jump_align_max_skip macros.
11404         * targhooks.h (default_loop_align_max_skip): Remove.
11405         (default_label_align_max_skip): Likewise.
11406         (default_jump_align_max_skip): Likewise.
11407         (default_label_align_after_barrier_max_skip): Remove.
11408         * toplev.c (read_log_maxskip): Use ::normalize function.
11409         (parse_N_M): Remove not used argument and also call ::normalize.
11410         (parse_alignment_opts): Do not pass unused arguments.
11411         * varasm.c (assemble_start_function): Use directly align_functions
11412         instead of removed macros.
11413         * system.h: Do not poison removed macros.
11415 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
11417         PR middle-end/86539
11418         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
11419         and cond temporaries don't have reference type if iterator has
11420         pointer type.  For init use &for_pre_body instead of pre_p if
11421         for_pre_body is non-empty.
11423 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
11425         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
11426         double-double modes to SFmode directly directly.
11427         (trunc<mode>sf2_fprs): Delete.
11429 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
11431         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
11432         for conversions between IFmode and the decimal floating point modes.
11433         (init_float128_ieee): Use the correct names for conversions between
11434         KFmode and the decimal floating point modes.
11436 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
11438         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
11439         for the conversions between TDmode and IFmode.
11440         (init_float128_ieee): Use more correct names for the conversions
11441         between TDmode and KFmode.
11443 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
11445         PR tree-optimization/86526
11446         * builtins.c (expand_builtin_memcmp): Formatting fixes.
11447         (inline_expand_builtin_string_cmp): Likewise.
11448         (inline_string_cmp): Likewise.  Use c_readstr instead of
11449         builtin_memcpy_read_str.  Add unit_mode temporary.
11451 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11453         PR middle-end/86528
11454         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
11455         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
11457 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11459         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
11460         Alphabetize prototypes of built-in functions, separating out
11461         built-in functions that are listed in this section but should be
11462         described elsewhere.
11464 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
11466         PR target/86511
11467         * expmed.c (emit_store_flag): Do not emit setcc followed by a
11468         conditional move when trapping comparison was split to a
11469         non-trapping one (and vice versa).
11471 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
11473         * config/s390/s390.c (s390_function_profiler): Generate nops
11474         instead of profiler call sequences.
11475         * config/s390/s390.opt: Add the new option.
11477 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
11479         * config/s390/s390.c (s390_function_profiler): Generate
11480         __mcount_loc section.
11481         * config/s390/s390.opt: Add the new option.
11483 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
11485         * common.opt: Add the new warning.
11486         * config/s390/s390.c (s390_function_profiler): Emit "brasl
11487         %r0,__fentry__" when -mfentry is specified.
11488         (s390_option_override_internal): Disallow -mfentry for 31-bit
11489         CPUs.
11490         * config/s390/s390.opt: Add the new option.
11492 2018-07-16  Richard Biener  <rguenther@suse.de>
11494         PR lto/86523
11495         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
11496         for function-local FUNCTION_DECL and RESULT_DECL immediately.
11498 2018-07-16  Martin Liska  <mliska@suse.cz>
11500         PR ipa/86529
11501         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
11502         to ::get_create.
11504 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
11506         * config/arc/arcHS.md: Update ARCHS scheduling rules.
11508 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
11510         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
11511         for ARCHS4x.
11512         * config/arc/arc-cpus.def (hs4x): New cpu.
11513         (hs4xd): Likewise.
11514         * config/arc/arc-tables.opt: Regenerate.
11515         * config/arc/arc.c (arc_sched_issue_rate): New function.
11516         (TARGET_SCHED_ISSUE_RATE): Define.
11517         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
11518         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
11519         fpu_cvt.
11520         (attr tune): Add ARCHS4x tune values.
11521         (attr tune_dspmpy): Define.
11522         (*tst): Correct instruction type.
11523         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
11524         * config/arc/arcHS4x.md: New file.
11525         * config/arc/fpu.md: Update instruction type attributes.
11526         * config/arc/t-multilib: Regenerate.
11528 2018-07-16  Tom de Vries  <tdevries@suse.de>
11530         PR debug/86455
11531         * var-tracking.c (vt_initialize): Fix pre_dec handling.
11533 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
11535         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
11536         early clobber.
11538 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
11540         PR tree-optimization/86514
11541         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
11542         conversion to a boolean type from a type with greater precision.
11544 2018-07-16  Tom de Vries  <tdevries@suse.de>
11546         * var-tracking.c (vt_initialize): Print adjusted insn slim if
11547         dump_flags request TDF_SLIM.
11549 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
11551         * fold-const.c (int_const_binop_1): Abstract...
11552         (wide_int_binop): ...wide int code here.
11553         (poly_int_binop): ...poly int code here.
11554         Abstract the rest of int_const_binop_1 into int_const_binop.
11555         * fold-const.h (wide_int_binop): New.
11556         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
11557         Remove useless PLUS/MINUS_EXPR case.
11558         (zero_nonzero_bits_from_vr): Move wide int code...
11559         (zero_nonzero_bits_from_bounds): ...here.
11560         (extract_range_from_binary_expr_1): Move mask optimization code...
11561         (range_easy_mask_min_max): ...here.
11562         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
11563         (range_easy_mask_min_max): New.
11565 2018-07-15  Jeff Law  <law@redhat.com>
11567         PR target/85993
11568         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
11569         block.
11571 2018-07-14  Jim Wilson  <jimw@sifive.com>
11573         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
11575 2018-07-14  Paul Koning  <ni1d@arrl.net>
11577         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
11579 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
11581         * lto-streamer-out.c (copy_function_or_variable): Dump info about
11582         copying section.
11584 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
11585             Steve Munroe  <munroesj52@gmail.com>
11587         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
11588         (_mm_andnot_si128): Likewise.
11589         (_mm_or_si128): Likewise.
11590         (_mm_xor_si128): Likewise.
11592 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
11594         PR middle-end/78809
11595         * builtins.c (expand_builtin_memcmp): Inline the calls first
11596         when result_eq is false.
11597         (expand_builtin_strcmp): Inline the calls first.
11598         (expand_builtin_strncmp): Likewise.
11599         (inline_string_cmp): New routine. Expand a string compare
11600         call by using a sequence of char comparison.
11601         (inline_expand_builtin_string_cmp): New routine. Inline expansion
11602         a call to str(n)cmp/memcmp.
11603         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
11604         New option.
11605         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
11607 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
11609         * config/arm/driver-arm.c: Include arm-native.h.
11610         (host_detect_local_cpu): Use auto-generated data tables.
11611         (vendors, arm_cpu_table): Delete.  Move part information to ...
11612         * config/arm/arm-cpus.in: ... here.
11613         * config/arm/parsecpu.awk (gen_native): New function.
11614         (vendor, part): New CPU fields.
11615         (END): Add support for building the native CPU detection tables.
11616         * config/arm/t-arm (arm-native.h): Add build rule.
11617         (driver-arm.o): Add dependency on arm-native.h.
11619 2018-07-13  Richard Biener  <rguenther@suse.de>
11621         PR middle-end/85974
11622         * match.pd (addr1 - addr2): Allow either of the operand to
11623         have a conversion.
11625 2018-07-13  Tom de Vries  <tdevries@suse.de>
11627         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
11628         in remap_ssa_name.
11630 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
11632         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
11633         arrays instead of numbered variables.
11635 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
11637         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
11638         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
11640 2018-07-13  Richard Biener  <rguenther@suse.de>
11642         PR debug/86452
11643         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
11644         instead of get_context_die.
11646 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
11647             Richard Biener  <rguenther@suse.de>
11649         PR middle-end/86489
11650         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
11651         that the loop latch destination where phi is defined.
11653 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
11655         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
11656         (riscv_expand_epilogue): Add assertion to check interrupt mode.
11657         (riscv_set_current_function): Extract getting interrupt type to new
11658         function.
11659         (riscv_get_interrupt_type): New function.
11660         (riscv_merge_decl_attributes): New function, checking interrupt type is
11661         same.
11662         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
11664 2018-07-12  Paul Koning  <ni1d@arrl.net>
11666         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
11667         directive.
11669 2018-07-12  Paul Koning  <ni1d@arrl.net>
11671         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
11672         zero reference, add doloop_end instead.
11673         * doc/md.texi (decrement_and_branch_until_zero): Remove.
11674         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
11675         detail for doloop_end.
11677 2018-07-12  Martin Sebor  <msebor@redhat.com>
11679         PR c/86453
11680         * attribs.c (decl_attributes): Reject conflicting attributes before
11681         calling attribute handlers.
11683 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
11685         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
11686          parameter.
11687         (gcc::dump_manager::get_dump_file_name): likewise.
11688         (dump_begin): Likewise.
11689         * dumpfile.h (dump_begin): Update prototype.
11690         (gcc::dump_manager::get_dump_file_name,
11691         gcc::dump_manager::get_dump_file_name): Update prototype.
11693 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
11695         * internal-fn.h (vectorizable_internal_fn_p): New function.
11696         * tree-vect-slp.c (compatible_calls_p): Likewise.
11697         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
11698         to internal functions.
11699         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
11701 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
11703         * fold-const.h (inverse_conditions_p): Declare.
11704         * fold-const.c (inverse_conditions_p): New function.
11705         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
11706         that test the inverse condition of a conditional internal function.
11707         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
11708         * internal-fn.c (internal_fn_mask_index): Handle conditional
11709         internal functions.
11710         (vectorized_internal_fn_supported_p): New function.
11711         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
11712         (any_pred_load_store): Replace with...
11713         (need_to_predicate): ...this new variable.
11714         (redundant_ssa_names): New variable.
11715         (ifcvt_can_use_mask_load_store): Move initial checks to...
11716         (ifcvt_can_predicate): ...this new function.  Handle tree codes
11717         for which a conditional internal function exists.
11718         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
11719         instead of ifcvt_can_use_mask_load_store.  Update after variable
11720         name change.
11721         (predicate_load_or_store): New function, split out from
11722         predicate_mem_writes.
11723         (check_redundant_cond_expr): New function.
11724         (value_available_p): Likewise.
11725         (predicate_rhs_code): Likewise.
11726         (predicate_mem_writes): Rename to...
11727         (predicate_statements): ...this.  Use predicate_load_or_store
11728         and predicate_rhs_code.
11729         (combine_blocks, tree_if_conversion): Update after above name changes.
11730         (ifcvt_local_dce): Handle redundant_ssa_names.
11731         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
11732         general conditional functions.
11733         * tree-vect-stmts.c (vectorizable_call): Likewise.
11735 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
11736             Alan Hayward  <alan.hayward@arm.com>
11737             David Sherwood  <david.sherwood@arm.com>
11739         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
11740         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
11741         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
11742         plus and minus and convert them into IFN_COND_FMA-based sequences.
11743         (convert_mult_to_fma): Handle conditional plus and minus.
11745 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
11747         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
11748         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
11749         (cond_fnms_optab): New optabs.
11750         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
11751         internal functions.
11752         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
11753         * internal-fn.h (get_conditional_internal_fn): Declare.
11754         (get_unconditional_internal_fn): Likewise.
11755         * internal-fn.c (cond_ternary_direct): New macro.
11756         (expand_cond_ternary_optab_fn): Likewise.
11757         (direct_cond_ternary_optab_supported_p): Likewise.
11758         (FOR_EACH_COND_FN_PAIR): Likewise.
11759         (get_conditional_internal_fn): New function.
11760         (get_unconditional_internal_fn): Likewise.
11761         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
11762         (gimple_match_op::gimple_match_op): Add a new overload for 5
11763         operands.
11764         (gimple_match_op::set_op): Likewise.
11765         (gimple_resimplify5): Declare.
11766         * genmatch.c (decision_tree::gen): Generate simplifications for
11767         5 operands.
11768         * gimple-match-head.c (gimple_simplify): Define an overload for
11769         5 operands.  Handle calls with 5 arguments in the top-level overload.
11770         (convert_conditional_op): Handle conversions from unconditional
11771         internal functions to conditional ones.
11772         (gimple_resimplify5): New function.
11773         (build_call_internal): Pass a fifth operand.
11774         (maybe_push_res_to_seq): Likewise.
11775         (try_conditional_simplification): Try converting conditional
11776         internal functions to unconditional internal functions.
11777         Handle 3-operand unconditional forms.
11778         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
11779         Define ternary equivalents of the current rules for binary conditional
11780         internal functions.
11781         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
11782         ternary operations.
11783         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
11784         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
11785         (optab): Handle them.
11786         (SVE_COND_FP_TERNARY): New int iterator.
11787         (sve_fmla_op, sve_fmad_op): New int attributes.
11788         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
11789         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
11790         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
11792 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
11794         * target.def (preferred_else_value): New target hook.
11795         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
11796         * doc/tm.texi: Regenerate.
11797         * targhooks.h (default_preferred_else_value): Declare.
11798         * targhooks.c (default_preferred_else_value): New function.
11799         * internal-fn.h (conditional_internal_fn_code): Declare.
11800         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
11801         (get_conditional_internal_fn): Use it.
11802         (conditional_internal_fn_code): New function.
11803         * gimple-match.h (gimple_match_cond): New struct.
11804         (gimple_match_op): Add a cond member function.
11805         (gimple_match_op::gimple_match_op): Update all forms to take a
11806         gimple_match_cond.
11807         * genmatch.c (expr::gen_transform): Use the same condition as res_op
11808         for the suboperation, but don't specify a particular else_value.
11809         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
11810         (visit_nary_op, visit_reference_op_load): Pass
11811         gimple_match_cond::UNCOND to the gimple_match_op constructor.
11812         * gimple-match-head.c: Include tree-eh.h
11813         (convert_conditional_op): New function.
11814         (maybe_resimplify_conditional_op): Likewise.
11815         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
11816         (gimple_resimplify2): Likewise.
11817         (gimple_resimplify3): Likewise.
11818         (gimple_resimplify4): Likewise.
11819         (maybe_push_res_to_seq): Return null for conditional operations.
11820         (try_conditional_simplification): New function.
11821         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
11822         constructor.
11823         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
11824         IFN_COND_* call.
11825         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
11826         function.
11827         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
11829 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
11831         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
11832         DECL_FCONTEXT
11833         (hash_tree): Do not hash DECL_FCONTEXT
11834         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
11835         Do not stream DECL_FCONTEXT.
11836         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
11837         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
11839 2018-07-12  Richard Biener  <rguenther@suse.de>
11841         PR debug/86462
11842         * dwarf2out.c (gen_block_die): Only output blocks when they have
11843         at least one !DECL_IGNORED_P variable.
11845 2018-07-12  Richard Biener  <rguenther@suse.de>
11847         PR target/84829
11848         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
11849         Remove -mieee-fp handling.
11851 2018-07-12  Richard Biener  <rguenther@suse.de>
11853         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
11854         left-over from last patch.
11856 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
11858         PR tree-optimization/86492
11859         * gimple-ssa-store-merging.c
11860         (imm_store_chain_info::coalesce_immediate_stores): Call
11861         check_no_overlap even for the merge_overlapping case.  Formatting fix.
11863 2018-07-12  Richard Biener  <rguenther@suse.de>
11865         PR middle-end/86479
11866         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
11867         move possibly trapping operations into the conditional.
11869 2018-07-12  Richard Biener  <rguenther@suse.de>
11871         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
11872         (vn_lookup_simplify_result): Remove recursion limit applied
11873         here.
11874         (vn_nary_build_or_lookup_1): Adjust.
11875         (try_to_simplify): Likewise.
11876         * gimple-match-head.c (gimple_resimplify1): Instead apply one
11877         here.
11878         (gimple_resimplify2): Likewise.
11879         (gimple_resimplify3): Likewise.
11880         (gimple_resimplify4): Likewise.
11882 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
11884         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
11885         Use __mmask64 type instead of __mmask8 for __M argument.
11886         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
11887         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
11888         __U argument.
11889         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
11890         __mmask16 for __M argument.
11891         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
11892         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
11893         to __mmask16 instead of __mmask8.
11894         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
11895         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
11896         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
11897         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
11898         instead of __mmask16 for __U argument.
11899         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
11900         __mmask16 instead of __mmask8 for __U argument.
11901         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
11902         __U argument.
11903         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
11904         __mmask16.
11905         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
11906         argument.
11907         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
11908         __U argument.
11909         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
11910         __mmask16.
11911         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
11912         of __mmask16.
11913         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
11914         __U argument.
11915         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
11916         __U argument.
11917         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
11918         __U argument.
11919         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
11920         __U argument.
11921         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
11922         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
11923         return type as well as __M argument type and all casts from __mmask8
11924         to __mmask32.
11925         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
11926         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
11927         return type as well as __M argument type and all casts from __mmask8
11928         to __mmask16.
11929         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
11930         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
11931         return type as well as __M argument type and all casts from __mmask8
11932         to __mmask32.
11933         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
11934         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
11935         return type as well as __M argument type and all casts from __mmask8
11936         to __mmask16.
11937         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
11938         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
11939         __mmask16.
11941 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
11943         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
11944         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
11945         for __U argument.
11947 2018-07-11  Paul Koning  <ni1d@arrl.net>
11949         * doc/md.texi (define_subst): Document how multiple occurrences of
11950         the same argument in the replacement pattern are handled.
11952 2018-07-11  Paul Koning  <ni1d@arrl.net>
11954         * doc/extend.texi (Common Variable Attributes): Move "mode" into
11955         alphabetical order.
11956         (Common Type Attributes): Add "mode" attribute.
11958 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
11960         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
11961         stream DECL_ORIGINAL_TYPE.
11962         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
11963         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
11964         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
11965         Do not walk original type.
11966         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
11967         external decls.
11968         (write_ts_decl_non_common_tree_pointers): Do not stream
11969         DECL_ORIGINAL_TYPE
11970         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
11971         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
11973 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
11975         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
11976         thread twice from the same starting edge.
11978 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
11980         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
11981         * gimple.c (gimple_call_nonnull_result_p): ...here...
11982         (gimple_call_nonnull_arg): ...and here.
11983         * gimple.h (gimple_call_nonnull_result_p): New.
11984         (gimple_call_nonnull_arg): New.
11986 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
11988         * config/arm/arm-cpus.in: Move information from fpu field of each
11989         cpu definition to the isa field.
11990         * config/arm/parsecpu.awk (fpu): Delete match rule.
11991         (gen_comm_data): Don't add bits from the CPU's FPU entry.
11993 2018-07-11  Richard Biener  <rguenther@suse.de>
11995         PR debug/86457
11996         * dwarf2out.c (init_sections_and_labels): Use
11997         output_asm_line_debug_info consistently.
11998         (dwarf2out_early_finish): Likewise.
11999         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
12000         type units.
12002 2018-07-11  Richard Biener  <rguenther@suse.de>
12004         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
12005         Rework father_bb setting in a way to avoid propagating constants
12006         multiple times on a loop body.
12008 2018-07-10  Mark Wielaard  <mark@klomp.org>
12010         PR debug/86459
12011         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
12013 2018-07-10  Richard Biener  <rguenther@suse.de>
12015         * hash-map.h (hash_map::iterator::operator*): Return
12016         references to key and value.
12018 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
12020         PR c++/86443
12021         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
12022         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
12023         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
12024         (gimplify_omp_for): For composite loops, move outer
12025         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
12026         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
12027         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
12028         TREE_LIST for both the original class iterator and the "last" helper
12029         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
12030         loop, remember has_decl_expr from outer composite loops for the
12031         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
12033 2018-07-09  Martin Sebor  <msebor@redhat.com>
12035         PR middle-end/77357
12036         PR middle-end/86428
12037         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
12038         accessing implicitly initialized array elements.
12039         * expr.c (string_constant): Handle string initializers of
12040         character arrays within aggregates.
12041         * gimple-fold.c (fold_array_ctor_reference): Add argument.
12042         Store element offset.  As a special case, handle zero size.
12043         (fold_nonarray_ctor_reference): Same.
12044         (fold_ctor_reference): Add argument.  Store subobject offset.
12045         * gimple-fold.h (fold_ctor_reference): Add argument.
12047 2018-07-09  Paul Koning  <ni1d@arrl.net>
12049         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
12050         (pdp11_insn_cost): New function.
12051         (pdp11_md_asm_adjust): New function.
12052         (TARGET_INVALID_WITHIN_DOLOOP): Define.
12053         (pdp11_rtx_costs): Update to match machine better.
12054         (output_addr_const_pdp11): Correct format mismatch warnings.
12055         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
12056         * config/pdp11/pdp11.md: General change to add base_cost and/or
12057         length attributes for use by new pdp11_insn_cost function.
12058         (MIN_BRANCH): Correct definition.
12059         (MIN_SOB): Ditto.
12060         (doloop_end): Use standard pattern name for looping pattern.
12061         (doloop_end_nocc): New.
12062         (movsf): Add another constraint alternative.
12063         (zero_extendqihi2): Add constraint alternatives for not in place
12064         extend.
12065         (zero_extendhisi2): Remove.
12066         (shift patterns): Add CC handling variants.
12067         (bswaphi2): New.
12068         (bswapsi2): New.
12069         (rothi3): New.
12070         (define_peephole2): New peephole to recognize mov that sets CC for
12071         subsequent test.
12073 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12075         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
12076         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
12077         wi::add.
12079 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
12081         PR c/86420
12082         * real.c (real_nextafter): Return true if result is denormal.
12084 2018-07-09  Martin Liska  <mliska@suse.cz>
12086         * common.opt: Add back wrongly removed attribute.
12088 2018-07-09  Richard Biener  <rguenther@suse.de>
12090         PR debug/86413
12091         * dwarf2out.c (gen_block_die): For an early generated DIE
12092         always output high/low PC attributes.
12094 2018-07-09  Tom de Vries  <tdevries@suse.de>
12096         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
12097         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
12098         onto VAR_DECL with abstract origin.
12100 2018-07-07  Jim Wilson  <jimw@sifive.com>
12102         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
12104 2018-07-07  Tom de Vries  <tdevries@suse.de>
12106         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
12108 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
12110         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
12111         overflow_type.
12112         (combine_bound): Use wide-int overflow calculation instead of
12113         rolling our own.
12114         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
12115         overflow_type.
12116         * fold-const.c (int_const_binop_2): Same.
12117         (extract_muldiv_1): Same.
12118         (fold_div_compare): Same.
12119         (fold_abs_const): Same.
12120         * match.pd: Same.
12121         * poly-int.h (add): Same.
12122         (sub): Same.
12123         (neg): Same.
12124         (mul): Same.
12125         * predict.c (predict_iv_comparison): Same.
12126         * profile-count.c (slow_safe_scale_64bit): Same.
12127         * simplify-rtx.c (simplify_const_binary_operation): Same.
12128         * tree-chrec.c (tree_fold_binomial): Same.
12129         * tree-data-ref.c (split_constant_offset_1): Same.
12130         * tree-if-conv.c (idx_within_array_bound): Same.
12131         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
12132         * tree-ssa-phiopt.c (minmax_replacement): Same.
12133         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
12134         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
12135         * vr-values.c (vr_values::adjust_range_with_scev): Same.
12136         * wide-int.cc (wi::add_large): Same.
12137         (wi::mul_internal): Same.
12138         (wi::sub_large): Same.
12139         (wi::divmod_internal): Same.
12140         * wide-int.h: Change overflow type to overflow_type for neg, add,
12141         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
12142         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
12143         mul_internal, divmod_internal.
12144         (overflow_type): New enum.
12145         (accumulate_overflow): New.
12147 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
12149         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
12150         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
12152 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
12154         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
12155         argument is checked for zero before entering loop, avoid checking again.
12157 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
12159         * gimplify.h (generic_expr_could_trap_p): Set as global function.
12160         * gimplify.h (generic_expr_could_trap_p): Likwise.
12161         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
12163 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
12165         PR tree-optimization/86401
12166         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
12167         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
12168         (fold_bit_and_mask): ... here.  New helper function for match.pd.
12169         * fold-const.h (fold_bit_and_mask): Declare.
12170         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
12172 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
12174         PR target/86324
12175         * target.def (translate_mode_attribute): New hook.
12176         * targhooks.h (default_translate_mode_attribute): Declare.
12177         * targhooks.c (default_translate_mode_attribute): New function.
12178         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
12179         * doc/tm.texi: Regenerate.
12180         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
12181         (rs6000_translate_mode_attribute): New function.
12183 2018-07-06  Paul Koning  <ni1d@arrl.net>
12185         * doc/md.texi (define_split): Document DONE and FAIL.
12186         (define_peephole2): Ditto.
12188 2018-07-05  Jeff Law  <law@redhat.com>
12190         PR tree-optimization/86010
12191         * tree-ssa-dse.c (compute_trims): More aggressively trim at
12192         both the head and tail of mem* and str* calls.
12194 2018-07-05  Jim Wilson  <jimw@sifive.com>
12196         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
12198 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
12200         * config/aarch64/aarch64-simd.md: correct flags text for
12201         MIN_EXPR replacement.
12203 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
12205         * configure: Regenerated.
12207 2018-07-05  Carl Love  <cel@us.ibm.com>
12209         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
12210         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
12211         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
12212         VSX_BUILTIN_DOUBLEL_V4SF.
12214 2018-07-05  Martin Sebor  <msebor@redhat.com>
12216         PR c++/86400
12217         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
12218         than its domain to compute its the upper bound of a char array.
12220 2018-07-05  Nathan Sidwell  <nathan@acm.org>
12222         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
12223         * doc/cpp.texi: Update comment.
12224         * doc/tm.texi: Rebuilt.
12225         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
12226         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
12227         * doc/extend.texi (Backwards Compatibility): Clarify it is system
12228         headers affected by extern "C".
12229         * system.h: Poison NO_IMPLICIT_EXTERN_C.
12230         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
12231         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
12232         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
12233         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
12234         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
12235         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
12236         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
12237         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
12238         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
12239         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
12240         config/sparc/sp64-elf.h, config/spu/spu.h,
12241         config/stormy16/stormy16.h, config/v850/v850.h,
12242         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
12243         define NO_IMPLICIT_EXTERN_C.
12244         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
12246 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
12248         PR target/84711
12249         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
12250         instead of GET_MODE_SIZE when comparing Units.
12252 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
12254         PR target/84711
12255         * rtlanal.c (set_noop_p): Constrain on mode change,
12256         include hard-reg-set.h
12258 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
12260         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
12262 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
12264         Revert
12265         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
12267         PR sanitizer/84250
12268         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
12269         libasan.
12270         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
12272 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
12274         PR sanitizer/84250
12275         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
12276         libasan.
12277         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
12279 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
12281         PR middle-end/86380
12282         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
12284 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
12286         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
12287         neg_*_op* variables.
12289 2018-07-04  Martin Liska  <mliska@suse.cz>
12291         * tree-switch-conversion.c: Define
12292         max_ratio_for_speed and max_ratio_for_size constants.
12294 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
12295             Martin Liska  <mliska@suse.cz>
12297         PR middle-end/66240
12298         PR target/45996
12299         PR c/84100
12300         * common.opt: Rename align options with 'str_' prefix.
12301         * common/config/i386/i386-common.c (set_malign_value): New
12302         function.
12303         (ix86_handle_option): Use it to set -falign-* options/
12304         * config/aarch64/aarch64-protos.h (struct tune_params): Change
12305         type from int to string.
12306         * config/aarch64/aarch64.c: Update default values from int
12307         to string.
12308         * config/alpha/alpha.c (alpha_override_options_after_change):
12309         Likewise.
12310         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
12311         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12312         max skip conditionally.
12313         * config/i386/freebsd.h (SUBALIGN_LOG): New.
12314         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12315         max skip conditionally.
12316         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12317         max skip conditionally.
12318         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
12319         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12320         max skip conditionally.
12321         * config/i386/i386.c (struct ptt): Change type from int to
12322         string.
12323         (ix86_default_align): Set default values.
12324         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
12325         max skip conditionally.
12326         * config/i386/iamcu.h (SUBALIGN_LOG): New.
12327         (ASM_OUTPUT_MAX_SKIP_ALIGN):
12328         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
12329         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12330         max skip conditionally.
12331         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
12332         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
12333         * config/i386/x86-64.h (SUBALIGN_LOG): New.
12334         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
12335         max skip conditionally.
12336         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12337         * config/ia64/ia64.c (ia64_option_override): Set default values
12338         for alignment options.
12339         * config/m68k/m68k.c: Handle new str_align_* options.
12340         * config/mips/mips.c (mips_set_compression_mode): Change
12341         type of constants.
12342         (mips_option_override): Set default values for options.
12343         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
12344         Likewise.
12345         * config/rs6000/rs6000.c (rs6000_option_override_internal):
12346         Likewise.
12347         * config/rx/rx.c (rx_option_override): Likewise.
12348         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
12349         (LABEL_ALIGN): Use align_labels_log.
12350         (LOOP_ALIGN): Use align_loops_align.
12351         * config/s390/s390.c (s390_asm_output_function_label): Use new
12352         macros.
12353         * config/sh/sh.c (sh_override_options_after_change):
12354         Change type of constants.
12355         * config/spu/spu.c (spu_sched_init): Likewise.
12356         * config/sparc/sparc.c (sparc_option_override): Set default
12357         values for options.
12358         * config/visium/visium.c (visium_option_override): Likewise.
12359         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
12360         emit p2align format with last argument if it's not needed.
12361         * doc/invoke.texi: Document extended format of -falign-*.
12362         * final.c: Use align_labels alignment.
12363         * flags.h (struct target_flag_state): Change type to use
12364         align_flags.
12365         (struct align_flags_tuple): New.
12366         (struct align_flags): Likewise.
12367         (align_loops_log): Redefine macro to use new types.
12368         (align_loops_max_skip): Redefine macro to use new types.
12369         (align_jumps_log): Redefine macro to use new types.
12370         (align_jumps_max_skip): Redefine macro to use new types.
12371         (align_labels_log): Redefine macro to use new types.
12372         (align_labels_max_skip): Redefine macro to use new types.
12373         (align_functions_log): Redefine macro to use new types.
12374         (align_loops): Redefine macro to use new types.
12375         (align_jumps): Redefine macro to use new types.
12376         (align_labels): Redefine macro to use new types.
12377         (align_functions): Redefine macro to use new types.
12378         (align_functions_max_skip): Redefine macro to use new types.
12379         (align_loops_value): New macro.
12380         (align_jumps_value): New macro.
12381         (align_labels_value): New macro.
12382         (align_functions_value): New macro.
12383         * function.c (invoke_set_current_function_hook): Propagate
12384         alignment values from flags to global variables default in
12385         topleev.h.
12386         * ipa-icf.c (sem_function::equals_wpa): Use
12387         cl_optimization_option_eq instead of memcmp.
12388         * lto-streamer.h (cl_optimization_stream_out): Support streaming
12389         of string types.
12390         (cl_optimization_stream_in): Likewise.
12391         * optc-save-gen.awk: Support strings in cl_optimization.
12392         * opth-gen.awk: Likewise.
12393         * opts.c (finish_options): Remove error checking of invalid
12394         value ranges.
12395         (MAX_CODE_ALIGN): Remove.
12396         (MAX_CODE_ALIGN_VALUE): Likewise.
12397         (parse_and_check_align_values): New function.
12398         (check_alignment_argument): Likewise.
12399         (common_handle_option): Use check_alignment_argument.
12400         * opts.h (parse_and_check_align_values): Declare.
12401         * toplev.c (init_alignments): Remove.
12402         (read_log_maxskip): New.
12403         (parse_N_M): Likewise.
12404         (parse_alignment_opts): Likewise.
12405         (backend_init_target): Remove usage of init_alignments.
12406         * toplev.h (parse_alignment_opts): Declare.
12407         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
12408         argument.
12409         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
12410         * tree.c (cl_option_hasher::equal): New.
12411         * varasm.c: Use new global macros.
12413 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
12415         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
12416         Use a simpler align directive also if MAXSKIP = ALIGN-1.
12417         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12418         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12419         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12420         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12421         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
12422         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
12423         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
12424         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12425         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12426         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12427         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12429 2018-07-04  Martin Liska  <mliska@suse.cz>
12430             Jonathan Wakely  <jwakely@redhat.com>
12432         * coverage.c: Use correct type.
12433         * doc/invoke.texi: Language correction.
12435 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
12437         PR target/85620
12438         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
12439         ENDBRANCH for non-tail call which may return via indirect branch.
12440         * doc/extend.texi: Document indirect_return attribute.
12442 2018-07-03  Martin Sebor  <msebor@redhat.com>
12444         PR tree-optimization/86274
12445         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
12446         precondition.
12447         (format_floating): Correct handling of infinities and NaNs.
12449 2018-07-03  Martin Sebor  <msebor@redhat.com>
12451         * print-tree.c (print_real_cst): New function.
12452         (print_node_brief): Call it.
12453         (print_node): Ditto.
12455 2018-07-03  Jeff Law  <law@redhat.com>
12457         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
12458         into a single pattern.
12460         * config/h8300/h8300.md (ors code_iterator): New.
12461         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
12462         a single pattern and single splitter.
12463         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
12464         (iorqi3_1, xorqi3_1): Likewise.
12465         (iorqi3, xorqi3 expanders): Similarly.
12467         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
12468         (movmd_internal) into a single pattern using the P mode iterator.
12469         (movmd splitters): Similarly.
12470         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
12471         (movsd splitters): Similarly.
12473         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
12474         ADDB, ADDW and ADDL into a single ADD attribute which selects the
12475         right table based on the size of the operand.
12476         * config/h8300/h8300.md (length_table): Corresponding changes. All
12477         references to "addb", "addw" and "addl" changed to "add".
12478         (btst patterns): Merge two variants into a single pattern.
12479         (tstqi, tsthi): Likewise.
12480         (addhi3_incdec, addsi3_incdec): Likewise.
12481         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
12482         (mulhi3, mulsi3): Likewise.
12483         (udivhi3, udivsi3): Likewise.
12484         (divhi3, divsi3): Likewise.
12485         (andorqi3, andorhi3, andorsi3): Likewise.
12487 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
12489         PR target/85694
12490         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
12491         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
12493 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12495         PR tree-optimization/85694
12496         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
12497         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
12498         UNSPEC_URHADD.
12499         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
12500         (<u>avg<mode>3_ceil): New patterns.
12502 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
12504         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
12505         scan-tree-dump directive.
12506         * gcc.dg/vect/slp-perm-2.c: Likewise.
12507         * gcc.dg/vect/slp-perm-3.c: Likewise.
12508         * gcc.dg/vect/slp-perm-5.c: Likewise.
12509         * gcc.dg/vect/slp-perm-6.c: Likewise.
12510         * gcc.dg/vect/slp-perm-7.c: Likewise.
12511         * gcc.dg/vect/slp-perm-8.c: Likewise.
12513 2018-07-03  Marek Polacek  <polacek@redhat.com>
12515         PR middle-end/86202
12516         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
12518 2018-07-03  Richard Biener  <rguenther@suse.de>
12520         PR ipa/86389
12521         * tree-ssa-structalias.c (find_func_clobbers): Properly
12522         handle indirect calls.
12524 2018-07-03  Jeff Law  <law@redhat.com>
12526         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
12527         (shifts): New code iterator.
12528         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
12529         expander.  Fix HImode handling on H8/SX.
12530         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
12531         (subqi3, subhi3, subsi3 expanders): Likewise.
12532         (andqi3, andhi3, andsi3 expanders): Likewise.
12533         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
12534         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
12535         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
12536         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
12537         (zero_extendqihi2, zero_extendqisi2): Likewise.
12538         (extendqihi2, extendqisi2): Likewise.
12539         (rotlqi3, rotlhi3, rotlsi3): Likewise.
12540         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
12541         (rotlqi3_1, rotlhi3_1): Likewise.
12542         (logicalhi3_sn, logicalsi3_sn): Likewise.
12543         (logicalhi3, logicalsi3): Likewise.
12545 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12547         * tree-vect-patterns.c (vect_recog_rotate_pattern)
12548         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
12549         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
12550         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
12551         type to append_pattern_def_seq instead of creating a stmt_vec_info
12552         directly.
12553         (build_mask_conversion): Likewise.  Remove vinfo argument.
12554         (vect_add_conversion_to_patterm): Likewise, renaming to...
12555         (vect_add_conversion_to_pattern): ...this.
12556         (vect_recog_mask_conversion_pattern): Update call to
12557         build_mask_conversion.  Pass the vector type to
12558         append_pattern_def_seq here too.
12559         (vect_recog_gather_scatter_pattern): Update call to
12560         vect_add_conversion_to_pattern.
12562 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12564         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
12565         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
12566         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
12567         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
12568         STMT_VINFO_PATTERN_DEF_SEQ to null here.
12569         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
12570         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
12571         append_pattern_def_seq instead of new_pattern_def_seq.
12572         (vect_recog_divmod_pattern): Do both of the above.
12573         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
12574         is null.
12576 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12578         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
12579         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
12580         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
12581         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
12582         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
12583         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
12584         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
12585         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
12586         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
12587         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
12588         parameter with a single stmt_vec_info.
12589         (vect_recog_func_ptr): Likewise.
12590         (vect_recog_gather_scatter_pattern): Likewise, folding in...
12591         (vect_try_gather_scatter_pattern): ...this.
12592         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
12593         the stmt_vec_info of the statement to be matched.  Don't clear
12594         STMT_VINFO_RELATED_STMT.
12595         (vect_pattern_recog): Update call accordingly.
12597 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12599         PR tree-optimization/85694
12600         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
12601         (uavgM3_ceil): Document new optabs.
12602         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
12603         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
12604         functions.
12605         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
12606         (savg_ceil_optab): New optabs.
12607         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
12608         (vect_vect_recog_func_ptrs): Add it.
12609         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
12610         constant directly from the associated lhs.
12612 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12614         * tree-vect-patterns.c (vect_split_statement): New function.
12615         (vect_convert_input): Use it to try to split an existing cast.
12617 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12619         * poly-int.h (print_hex): New function.
12620         * dumpfile.h (dump_dec, dump_hex): Declare.
12621         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
12622         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
12623         min_input_precision, operation_precision and operation_sign.
12624         * tree-vect-patterns.c (vect_get_range_info): New function.
12625         (vect_same_loop_or_bb_p, vect_single_imm_use)
12626         (vect_operation_fits_smaller_type): Delete.
12627         (vect_look_through_possible_promotion): Add an optional
12628         single_use_p parameter.
12629         (vect_recog_over_widening_pattern): Rewrite to use new
12630         stmt_vec_info infomration.  Handle one operation at a time.
12631         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
12632         (vect_truncatable_operation_p, vect_set_operation_type)
12633         (vect_set_min_input_precision): New functions.
12634         (vect_determine_min_output_precision_1): Likewise.
12635         (vect_determine_min_output_precision): Likewise.
12636         (vect_determine_precisions_from_range): Likewise.
12637         (vect_determine_precisions_from_users): Likewise.
12638         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
12639         (vect_vect_recog_func_ptrs): Put over_widening first.
12640         Add cast_forwprop.
12641         (vect_pattern_recog): Call vect_determine_precisions.
12643 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
12645         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
12646         statements that have been replaced by further pattern statements.
12647         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
12649 2018-07-03  Richard Biener  <rguenther@suse.de>
12651         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
12652         always set *dt.  Dump vectype in vectype overload.
12653         * dumpfile.h (dump_gimple_expr): New function.
12654         (dump_gimple_expr_loc): Likewise.
12655         * dumpfile.c (dump_gimple_expr): New function.
12656         (dump_gimple_expr_loc): Likewise.
12658 2018-07-02  Jeff Law  <law@redhat.com>
12660         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
12661         the H8/300, H8/300H and H8/S variants into a single pattern.
12662         (movhi_h8300, movqi_h8300hs): Similarly.
12663         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
12664         (QHI mode iterator): New.
12666         * config/h8300/h8300.md: Remove trailing whitespace.
12668 2018-07-02  Jim Wilson  <jimw@sifive.com>
12670         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
12671         instead of emit_insn for interrupt returns.
12672         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
12673         (riscv_sret, riscv_uret): Likewise.
12675 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
12677         * pretty-print.c (selftest::test_pp_format): Move save and restore
12678         of quotes to class auto_fix_quotes, and add an instance.
12679         * selftest.c: Include "intl.h".
12680         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
12681         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
12682         * selftest.h (selftest::auto_fix_quotes): New class.
12684 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
12686         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
12687         (aarch64_sve_prepare_conditional_op): Remove.
12688         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
12689         Allow aarch64_simd_reg_or_zero as select operand; remove
12690         the aarch64_sve_prepare_conditional_op call.
12691         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
12692         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
12693         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
12694         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
12695         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
12696         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
12697         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
12698         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
12699         and a splitters to match all of the *_any patterns.
12700         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
12702         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
12703         (SVE_COND_FP_BINARY_REV): Remove.
12704         (sve_int_op_rev, sve_fp_op_rev): New.
12705         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
12706         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
12707         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
12708         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
12709         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
12710         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
12711         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
12712         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
12713         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
12715         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
12716         Remove match_dup 1 from the inner unspec.
12717         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
12719         * config/aarch64/aarch64.md (movprfx): New attr.
12720         (length): Default movprfx to 8.
12721         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
12722         (*madd<SVE_I>, *msub<SVE_I): Likewise.
12723         (*<su>mul<SVE_I>3_highpart): Likewise.
12724         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
12725         (*v<ASHIFT><SVE_I>3): Likewise.
12726         (*<su><MAXMIN><SVE_I>3): Likewise.
12727         (*<su><MAXMIN><SVE_F>3): Likewise.
12728         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
12729         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
12730         (*div<SVE_F>4): Likewise.
12732 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
12734         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
12735         in dump string.
12737 2018-07-02  Richard Biener  <rguenther@suse.de>
12739         PR tree-optimization/86363
12740         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
12741         memset argument refers to a non-variable address.
12743 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
12745         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
12746         {PLUS,MINUS}_EXPR code to...
12747         (adjust_symbolic_bound): ...here,
12748         (combine_bound): ...here,
12749         (set_value_range_with_overflow): ...and here.
12751 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
12753         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
12754         code...
12755         (extract_range_from_abs_expr): ...here.
12757 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
12759         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
12760         -fno-omit-frame-pointer when not optimizing.
12762 2018-07-02  Martin Liska  <mliska@suse.cz>
12764         PR ipa/86279
12765         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
12766         (propagate_nothrow): Likewise.
12768 2018-07-02  Martin Liska  <mliska@suse.cz>
12770         PR ipa/86323
12771         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
12773 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
12775         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
12776         function in r262149, changing "loc" param from source_location to
12777         const dump_location_t &.
12778         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
12779         declaration, as above.
12781 2018-07-01  Paul Koning  <ni1d@arrl.net>
12783         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
12784         -munit-asm, -mgnu-asm, -mdec-asm.
12785         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
12786         (pdp11_output_labelref): New.
12787         (pdp11_output_def): New.
12788         (pdp11_output_addr_vec_elt): New.
12789         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
12790         %# and %@ format codes.
12791         (pdp11_option_override): New.
12792         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
12793         (pdp11_output_ident): New.
12794         (pdp11_asm_named_section): New.
12795         (pdp11_asm_init_sections): New.
12796         (pdp11_file_start): New.
12797         (pdp11_file_end): New.
12798         (output_ascii): Use .ascii/.asciz for -mdec-asm.
12799         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
12800         %o, like %c but octal.
12801         (pdp11_option_override): New.
12802         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
12803         -mdec-asm.
12804         (DATA_SECTION_ASM_OP): Ditto.
12805         (READONLY_DATA_SECTION_ASM_OP): New.
12806         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
12807         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
12808         (ASM_OUTPUT_LABELREF): Ditto.
12809         (ASM_OUTPUT_DEF): Ditto.
12810         (ASM_OUTPUT_EXTERNAL): New.
12811         (ASM_OUTPUT_SOURCE_FILENAME): New.
12812         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
12813         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
12814         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
12815         %# and %@ format codes.
12816         * config/pdp11/pdp11.opt (mgnu-asm): New.
12817         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
12818         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
12819         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
12821 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
12823         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
12824         dereferencing path[] beyond its length.
12825         (debug_path): New.
12826         (debug_all_paths): New.
12827         (rewire_first_differing_edge): New.
12828         (adjust_paths_after_duplication): New.
12829         (duplicate_thread_path): Call adjust_paths_after_duplication.
12830         Add new argument.
12831         (thread_through_all_blocks): Add new argument to
12832         duplicate_thread_path.
12834 2018-06-30  Jim Wilson  <jimw@sifive.com>
12836         * config/riscv/predicates.md (p2m1_shift_operand): New.
12837         (high_mask_shift_operand): New.
12838         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
12839         pattern using p2m1_shift_operand.
12840         (lshsi3_zero_extend_3+2): New combiner pattern using
12841         high_mask_shift_operand.
12843 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
12845         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
12846         split out from...
12847         (vect_recog_rotate_pattern): ...here.
12848         (vect_convert_input): Try to insert casts of invariants in the
12849         preheader.
12850         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
12851         preheader to be empty.
12853 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
12855         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
12856         vector type.  If given, install it in the new statement's
12857         STMT_VINFO_VECTYPE.
12858         (vect_element_precision): New function.
12859         (vect_unpromoted_value): New struct.
12860         (vect_unpromoted_value::vect_unpromoted_value): New function.
12861         (vect_unpromoted_value::set_op): Likewise.
12862         (vect_look_through_possible_promotion): Likewise.
12863         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
12864         (vect_widened_op_tree, vect_convert_input): Likewise.
12865         (vect_convert_inputs, vect_convert_output): Likewise.
12866         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
12867         to handle the optional cast of the multiplication result and
12868         vect_widened_op_tree to detect the widened multiplication itself.
12869         Do not require the input and output of promotion casts to have
12870         the same sign, but base the signedness of the operation on the
12871         input rather than the result.  If the pattern includes two
12872         promotions, check that those promotions have the same sign.
12873         Do not restrict the MULT_EXPR handling to a double-width result;
12874         handle quadruple-width results and wider.  Use vect_convert_inputs
12875         to convert the inputs to the common type.
12876         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
12877         to handle the optional cast of the ABS result.  Also allow a sign
12878         change or a sign extension between the ABS and MINUS.
12879         Use vect_widened_op_tree to detect the widened subtraction and use
12880         vect_convert_inputs to convert the inputs to the common type.
12881         (vect_handle_widen_op_by_const): Delete.
12882         (vect_recog_widen_op_pattern): New function.
12883         (vect_recog_widen_mult_pattern): Use it.
12884         (vect_recog_widen_shift_pattern): Likewise.
12885         (vect_recog_widen_sum_pattern): Use
12886         vect_look_through_possible_promotion to handle the promoted
12887         PLUS_EXPR operand.
12889 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
12891         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
12892         the containing gimple_seq *.
12893         * gimple-iterator.h (gsi_for_stmt): Declare it.
12894         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
12895         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
12896         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
12897         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
12898         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
12899         checks.
12900         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
12901         split out from...
12902         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
12903         statement being replaced is part of an existing pattern
12904         definition sequence, inserting the new pattern statements before
12905         the original one.
12906         (vect_pattern_recog_1): Don't return a bool.  If the statement
12907         is already part of a pattern, instead apply pattern matching
12908         to the pattern definition statements.  Don't clear the
12909         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
12910         (vect_pattern_recog): Don't break after the first match;
12911         continue processing the pattern definition statements instead.
12912         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
12914 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
12916         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12917         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
12918         (vect_recog_widen_sum_pattern): Use it.
12920 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
12922         * tree-vect-loop.c (vectorizable_reduction): Assert that the
12923         phi is not a pattern statement and has not been replaced by
12924         a pattern statement.
12925         * tree-vect-patterns.c (type_conversion_p): Don't check
12926         STMT_VINFO_IN_PATTERN_P.
12927         (vect_recog_vector_vector_shift_pattern): Likewise.
12928         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
12929         the pattern statement rather than the original statement; check
12930         directly for a WIDEN_MULT_EXPR here.
12931         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
12932         vect_is_simple_use to return the pattern statement rather
12933         than the original statement; use is_pattern_stmt_p to check
12934         for such a pattern statement.
12935         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
12936         to return the pattern statement rather than the original statement;
12937         don't do the same transformation here.
12938         (vect_is_simple_use): If the defining statement has been replaced
12939         by a pattern statement, return the pattern statement instead.
12940         Remove the corresponding (local) transformation from the vectype
12941         overload.
12943 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
12945         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
12946         end and default to null.
12947         * tree-vect-loop.c (vect_create_epilog_for_reduction)
12948         (vectorizable_reduction): Update calls accordingly, dropping the
12949         gimple ** argument if the passed-back statement isn't needed.
12950         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
12951         (vect_recog_rotate_pattern): Likewise.
12952         (vect_recog_mask_conversion_pattern): Likewise.
12953         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
12954         (vect_mask_constant_operand_p): Likewise.
12955         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
12956         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
12957         (get_group_load_store_type, get_load_store_type): Likewise.
12958         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
12959         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
12960         (vectorizable_conversion, vectorizable_assignment): Likewise.
12961         (vectorizable_shift, vectorizable_operation): Likewise.
12962         (vectorizable_store, vect_is_simple_cond): Likewise.
12963         (vectorizable_condition, vectorizable_comparison): Likewise.
12964         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
12965         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
12966         and move it to the end.  Cope with null def_stmt_outs.
12968 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12970         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
12972 2018-06-29  Jeff Law  <law@redhat.com>
12974         * config/v850/v850.c (v850_legitimate_address_p): Handle large
12975         displacements for TARGET_V850E2V3 and newer.
12976         (TARGET_LRA_P): Remove.  Defaults to LRA now.
12977         * config/v850/v850.md (sign23byte_load): Remove.
12978         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
12979         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
12981 2018-06-29  Martin Liska  <mliska@suse.cz>
12983         PR lto/85759
12984         * coverage.c (coverage_init): Mangle full path name.
12985         * doc/invoke.texi: Document the change.
12986         * gcov-io.c (mangle_path): New.
12987         * gcov-io.h (mangle_path): Likewise.
12988         * gcov.c (mangle_name): Use mangle_path for path mangling.
12990 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12992         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
12993         if starting source register is not even.
12995 2018-06-29  Martin Liska  <mliska@suse.cz>
12997         PR tree-optimization/86263
12998         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
12999         Make edge redirection.
13001 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
13003         * dumpfile.c (dump_loc): Add indentation based on scope depth.
13004         (dump_scope_depth): New variable.
13005         (get_dump_scope_depth): New function.
13006         (dump_begin_scope): New function.
13007         (dump_end_scope): New function.
13008         * dumpfile.h (get_dump_scope_depth): New declaration.
13009         (dump_begin_scope): New declaration.
13010         (dump_end_scope): New declaration.
13011         (class auto_dump_scope): New class.
13012         (AUTO_DUMP_SCOPE): New macro.
13013         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
13014         AUTO_DUMP_SCOPE.
13016 2018-06-29  Richard Biener  <rguenther@suse.de>
13018         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
13019         compute_all_dependences succeeds.
13020         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
13021         exceed --param loop-max-datarefs-for-datadeps.
13023 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
13025         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
13027 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
13029         PR target/86348
13030         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
13031         alternative 0 in preferred_for_speed attribute.
13033 2018-06-28  Paul Koning  <ni1d@arrl.net>
13035         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
13036         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
13037         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
13038         * config/pdp11/pdp11.md: Correct "length" attribute calculation
13039         for shift insn patterns.
13041 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
13043         * cgraph.c (cgraph_node::get_body): Replace assignments to
13044         "dump_file" with calls to set_dump_file.
13045         * dumpfile.c (alt_dump_file): Make static, and group with...
13046         (alt_flags): ...this definition.
13047         (dumps_are_enabled): New variable.
13048         (refresh_dumps_are_enabled): New function.
13049         (set_dump_file): New function.
13050         (set_alt_dump_file): New function.
13051         (gcc::dump_manager::dump_start): Replace assignments to
13052         "dump_file" and "alt_dump_file" with calls to set_dump_file and
13053         set_alt_dump_file.
13054         (gcc::dump_manager::dump_finish): Likewise.
13055         * dumpfile.h (alt_dump_file): Delete decl.
13056         (dumps_are_enabled): New variable decl.
13057         (set_dump_file): New function decl.
13058         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
13059         global.
13060         * tree-nested.c (lower_nested_functions): Replace assignments to
13061         "dump_file" with calls to set_dump_file.
13063 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
13065         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
13066         goto_locus of each outgoing edge of each basic block.
13068 2018-06-28  Richard Biener  <rguenther@suse.de>
13070         * dwarf2out.c (decl_scope_table): Remove.
13071         (push_decl_scope): Likewise.
13072         (pop_decl_scope): Likewise.
13073         (gen_type_die_for_member): Do not call push/pop_decl_scope.
13074         (gen_struct_or_union_type_die): Likewise.
13075         (gen_tagged_type_die): Likewise.
13076         (dwarf2out_init): Do not initialize decl_scope_table.
13077         (dwarf2out_c_finalize): Do not free it.
13079 2018-06-28  Richard Biener  <rguenther@suse.de>
13081         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
13082         deciding whether to not re-use a DIE.
13084 2018-06-28  Richard Biener  <rguenther@suse.de>
13086         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
13087         DW_AT_abstract_origin attribute.
13089 2018-06-28  Martin Liska  <mliska@suse.cz>
13091         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
13092         Use newly introduced constants.
13093         * tree-switch-conversion.h (struct jump_table_cluster):
13094         Define max_ratio_for_size and max_ratio_for_speed.
13096 2018-06-28  Martin Liska  <mliska@suse.cz>
13098         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
13099         Add new checking assert to catch invalid state.
13100         (jump_table_cluster::can_be_handled): Handle single case
13101         clusters.
13102         (jump_table_cluster::is_beneficial): Bail out for such case.
13103         (bit_test_cluster::find_bit_tests):
13104         Add new checking assert to catch invalid state.
13105         (bit_test_cluster::can_be_handled): Handle single case
13106         clusters.
13107         (bit_test_cluster::is_beneficial): Bail out for such case.
13108         (switch_decision_tree::analyze_switch_statement):
13109         Fix comment.
13111 2018-06-28  Martin Liska  <mliska@suse.cz>
13113         * common.opt: Introduce -completion option.
13114         * gcc.c (driver_handle_option): Handle it.
13115         (driver::main): Print completions if completion
13116         is set.
13117         * opt-suggestions.c (option_proposer::get_completions):
13118         New function.
13119         (option_proposer::suggest_completion): Likewise.
13120         (option_proposer::find_param_completions): Likewise.
13121         (verify_autocompletions): Likewise.
13122         (test_completion_valid_options): Likewise.
13123         (test_completion_valid_params): Likewise.
13124         (in_completion_p): Likewise.
13125         (empty_completion_p): Likewise.
13126         (test_completion_partial_match): Likewise.
13127         (test_completion_garbage): Likewise.
13128         (opt_proposer_c_tests): Likewise.
13129         * opt-suggestions.h: Declare new functions.
13130         * opts.c (common_handle_option): Handle OPT__completion_.
13131         * selftest-run-tests.c (selftest::run_tests): Add
13132         opt_proposer_c_tests.
13133         * selftest.c (assert_str_startswith): New.
13134         * selftest.h (assert_str_startswith): Likewise.
13135         (opt_proposer_c_tests): New.
13136         (ASSERT_STR_STARTSWITH): Likewise.
13138 2018-06-28  Martin Liska  <mliska@suse.cz>
13140         * Makefile.in: Add opt-suggestions.o.
13141         * gcc-main.c: Include opt-suggestions.h.
13142         * gcc.c (driver::driver): Likewise.
13143         (driver::~driver): Remove m_option_suggestions.
13144         (driver::build_option_suggestions): Moved to option_proposer.
13145         (driver::suggest_option): Likewise.
13146         (driver::handle_unrecognized_options): Use option_proposer.
13147         * gcc.h (class driver): Add new memver m_option_proposer.
13148         * opt-suggestions.c: New file.
13149         * opt-suggestions.h: New file.
13151 2018-06-28  Martin Liska  <mliska@suse.cz>
13153         * vec.h (class auto_string_vec): New (moved from auto_argvec).
13154         (auto_string_vec::~auto_string_vec): Likewise.
13156 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
13158         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
13159         prevent_decl_creation_for_types fields up and add reset_location field.
13160         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
13161         statement if id->reset_location is true.
13162         (copy_edges_for_bb): Do not set goto_locus on the new edges if
13163         id->reset_location is true.
13164         (copy_phis_for_bb): Force input_location on the arguments if
13165         id->reset_location is true.
13166         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
13167         is set on the function to be inlined.
13169 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
13171         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
13173 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
13175         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
13176         registers for Pmode.
13177         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
13178         hard registers for the clobbered pseudo.
13180 2018-06-27  Paul Koning  <ni1d@arrl.net>
13182         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
13183         mutually exclusive options.
13184         * config/pdp11/constraints.md (h): New constraint.
13185         (O): Update definition to match shift code generation.
13186         (D): New constraint.
13187         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
13188         (CCFP): Remove.
13189         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
13190         function.
13191         (output_jump): Change arguments.
13192         (pdp11_fixed_cc_regs): New function.
13193         (pdp11_cc_mode): Ditto.
13194         (pdp11_expand_shift): Ditto.
13195         (pdp11_assemble_shift): Ditto.
13196         (pdp11_small_shift): Ditto.
13197         (pdp11_branch_cost): Remove.
13198         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
13199         from output.
13200         (pdp11_register_move_cost): Update for CC registers.
13201         (pdp11_rtx_costs): Add case for LSHIFTRT.
13202         (pdp11_output_jump): Add CCNZ mode conditional branches.
13203         (notice_update_cc_on_set): Remove.
13204         (pdp11_cc_mode): New function.
13205         (simple_memory_operand): Correct pre/post decrement case.
13206         (no_side_effect_operand): New function.
13207         (pdp11_regno_reg_class): Add CC_REGS class.
13208         (pdp11_fixed_cc_regs): New function.
13209         (pdp11_small_shift): New function.
13210         (pdp11_expand_shift): New function to expand shift insns.
13211         (pdp11_assemble_shift): New function to output shifts.
13212         (pdp11_branch_cost): Remove.
13213         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
13214         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
13215         (WCHAR_TYPE): Ditto.
13216         (PTRDIFF_TYPE): Ditto.
13217         (ADJUST_INSN_LENGTH): New macro.
13218         (FIXED_REGISTERS): Add CC registers.
13219         (CALL_USED_REGISTERS): Ditto.
13220         (reg_class): Ditto.
13221         (REG_CLASS_NAMES): Ditto.
13222         (REG_CLASS_CONTENTS): Ditto.
13223         (SELECT_CC_MODE): Use new function.
13224         (TARGET_FLAGS_REGNUM): New macro.
13225         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
13226         (cc0_reg_rtx): Remove.
13227         (CC_STATUS_MDEP): Remove.
13228         (CC_STATUS_MDEFP_INIT): Remove.
13229         (CC_IN_FPU): Remove.
13230         (NOTICE_UPDATE_CC): Remove.
13231         (REGISTER_NAMES): Add CC registers.
13232         (BRANCH_COST): Change to constant 1.
13233         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
13234         handling.
13235         * config/pdp11/pdp11.opt (mbcopy): Remove.
13236         (mbcopy-builtin): Remove.
13237         (mbranch-cheap): Remove.
13238         (mbranch-expensive): Remove.
13239         * config/pdp11/predicates.md (expand_shift_operand): Update to
13240         match shift code generation.
13241         (ccnz_operator): New predicate.
13242         * doc/invoke.texi (PDP-11 Options): Remove deleted options
13243         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
13244         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
13245         exclusive options.
13246         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
13247         description of O constraint.
13249 2018-06-27  Jeff Law  <law@redhat.com>
13250             Austin Law  <austinklaw@gmail.com>
13252         * config/v850/v850.md (addsi3_set_flags): New pattern.
13253         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
13254         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
13255         (zero_extendhisi2_v850_set_flags): Likewise.
13256         (zero_extendqisi2_v850_set_flags): Likewise.
13257         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
13258         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
13259         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
13261         * config/v850/v850-protos.h (notice_update_cc): Remove.
13262         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
13263         (v850_print_operand): Handle 'D' and "d".
13264         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
13265         Add handling of arithmetic/logical operations compared against zero.
13266         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
13267         Do not look at v850_compare_op, instead get mode from last argument.
13268         (v850_gen_compare): Remove
13269         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
13270         after reload for prologue insns.
13271         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
13272         patterns.
13273         (construct_save_jarl): Likewise.
13274         (TARGET_FLAGS_REGNUM): Define.
13275         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
13276         (NOTICE_UPDATE_CC): Remove.
13277         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
13278         than cc0.  Conditionalize on reload_completed.
13279         (cmpsi_insn, setfcc_insn): Likewise.
13280         (tst1 splitter): Turn into define_and_split which sets the flags
13281         after reload.
13282         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
13283         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
13284         (cstoresf4, cstoredf4): Clobber the flags.
13285         (cmpsi, cmpsf, cmpdf): Remove expanders.
13286         (setf_insn): Remove pattern.
13287         (addsi3): Turn into define_and_split which clobbers the flags after
13288         reload and a suitable pattern (addsi3_clobber_flags) for use after
13289         reload.
13290         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
13291         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
13292         (ashrsi3, ashrsi3_v850e2): Likewise.
13293         (bins): Clobber the flags.
13294         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
13295         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
13296         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
13297         (call_value_internal_short, call_value_internal_long): Likewise.
13298         (callt_save_interrupt, callt_return_interrupt): Likewise.
13299         (save_interrupt, return_interrupt): Likewise.
13300         (callt_save_all_interrupt, save_all_interrupt): Likewise.
13301         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
13302         (restore_all_interrupt, _restore_all_interrupt): Likewise.
13303         (All FP comparisons): Only allow after reload has completed.
13304         (trfsr): Likewise.
13305         (divh, divhu): Tweak output template.
13306         (branch_z_normal, branch_z_invert): Remove
13307         (branch_nz_normal, branch_nz_invert): Likewise.
13308         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
13310         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
13311         * config/v850/v850.c (notice_update_cc): Remove.
13312         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
13313         (CC_NO_CARRY): Likewise.
13314         (NOTICE_UPDATE_CC): Define to nothing.
13315         * config/v850/v850.md: Remove block comment on cc0 handling
13316         Remove "cc" attribute from all patterns.  Remove cc_status handling
13317         from all patterns.  Minor formatting fixes.
13319 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13321         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
13322         (cortex-a76.cortex-a55): Likewise.
13323         * config/aarch64/aarch64-tune.md: Regenerate.
13324         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
13325         cortex-a76.cortex-a55.
13327 2018-06-27  Jeff Law  <law@redhat.com>
13329         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
13330         (MULTILIB_DIRNAMES): Similarly.
13332 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
13334         * gimple.h (gimple_return_retbnd): Delete.
13335         (gimple_return_set_retbnd): Likewise.
13336         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
13337         gimple_return_set_retbnd.
13338         * gimple-pretty-print.c (dump_gimple_return): Remove call to
13339         gimple_return_retbnd and adjust.
13340         * tree-inline.h (struct copy_body_data): Remove retbnd field.
13341         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
13342         Explicitly return NULL in a couple more cases.  Move assertion
13343         on debug statements and remove unreachable code.
13344         (reset_debug_binding): Do not test id->retbnd.
13345         (expand_call_inline): Do not set it.
13347 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
13349         * configure.ac: Add --disable-gcov option.
13350         * configure: Regenerate.
13351         * Makefile.in: Honour @enable_gcov@.
13352         * doc/install.texi: Document --disable-gcov.
13354 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13356         * config/arm/arm-cpus.in (cortex-a76): New entry.
13357         (cortex-a76.cortex-a55): Likewise.
13358         * config/arm/arm-tables.opt: Regenerate.
13359         * config/arm/arm-tune.md: Likewise.
13360         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
13361         * doc/invoke.texi (ARM Options): Document cortex-a76 and
13362         cortex-a76.cortex-a55.
13364 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
13366         PR target/85769
13367         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
13369 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
13371         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
13372         comment.
13373         (EPILOGUE_USES): Likewise.
13375 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
13377         * tree-inline.c (remap_location): New function extracted from...
13378         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
13379         (copy_phis_for_bb): ...here.  Call remap_location.
13380         (copy_cfg_body): Adjust call to copy_edges_for_bb.
13382 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
13384         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
13385         unaligned vsx for 16B memset.
13387 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
13389         PR target/86285
13390         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
13391         ieee128_float_type_node to long_double_type_node unless
13392         TARGET_LONG_DOUBLE_128 is set.
13394 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
13396         * cfgloop.c (get_loop_location): Convert return type from
13397         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
13398         by implicit construction from rtx_insn *, and using
13399         dump_user_location_t::from_function_decl for the fallback case.
13400         * cfgloop.h (get_loop_location): Convert return type from
13401         location_t to dump_user_location_t.
13402         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
13403         dump_printf_loc to pass in a dump_location_t rather than a
13404         location_t, via the gimple stmt.
13405         * coverage.c (get_coverage_counts): Update calls to
13406         dump_printf_loc to pass in dump_location_t rather than a
13407         location_t.
13408         * doc/optinfo.texi (Dump types): Convert example of
13409         dump_printf_loc from taking "locus" to taking "insn".  Update
13410         description of the "_loc" calls to cover dump_location_t.
13411         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
13412         "selftest.h".
13413         (dump_user_location_t::dump_user_location_t): New constructors,
13414         from gimple *stmt and rtx_insn *.
13415         (dump_user_location_t::from_function_decl): New function.
13416         (dump_loc): Make static.
13417         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
13418         const dump_location_t &.
13419         (dump_generic_expr_loc): Delete.
13420         (dump_printf_loc): Convert param "loc" from location_t to
13421         const dump_location_t &.
13422         (selftest::test_impl_location): New function.
13423         (selftest::dumpfile_c_tests): New function.
13424         * dumpfile.h: Include "profile-count.h".
13425         (class dump_user_location_t): New class.
13426         (struct dump_impl_location_t): New struct.
13427         (class dump_location_t): New class.
13428         (dump_printf_loc): Convert 2nd param from source_location to
13429         const dump_location_t &.
13430         (dump_generic_expr_loc): Delete.
13431         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
13432         const dump_location_t &.
13433         * gimple-fold.c (fold_gimple_assign): Update call to
13434         dump_printf_loc to pass in a dump_location_t rather than a
13435         location_t, via the gimple stmt.
13436         (gimple_fold_call): Likewise.
13437         * gimple-loop-interchange.cc
13438         (loop_cand::analyze_iloop_reduction_var): Update for change to
13439         check_reduction_path.
13440         (tree_loop_interchange::interchange): Update for change to
13441         find_loop_location.
13442         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
13443         change in return-type of find_loop_location.
13444         (graphite_regenerate_ast_isl): Likewise.
13445         * graphite-optimize-isl.c (optimize_isl): Likewise.
13446         * graphite.c (graphite_transform_loops): Likewise.
13447         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
13448         pass in a dump_location_t rather than a location_t, via the
13449         gimple stmt.
13450         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13451         * ipa.c (walk_polymorphic_call_targets): Likewise.
13452         * loop-unroll.c (report_unroll): Convert "locus" param from
13453         location_t to dump_location_t.
13454         (decide_unrolling): Update for change to get_loop_location's
13455         return type.
13456         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
13457         location_t to dump_user_location_t.
13458         (grid_find_single_omp_among_assignments_1): Updates calls to
13459         dump_printf_loc to pass in a dump_location_t rather than a
13460         location_t, via the gimple stmt.
13461         (grid_parallel_clauses_gridifiable): Convert "tloc" from
13462         location_t to dump_location_t.  Updates calls to dump_printf_loc
13463         to pass in a dump_location_t rather than a location_t, via the
13464         gimple stmt.
13465         (grid_inner_loop_gridifiable_p): Likewise.
13466         (grid_dist_follows_simple_pattern): Likewise.
13467         (grid_gfor_follows_tiling_pattern): Likewise.
13468         (grid_target_follows_gridifiable_pattern): Likewise.
13469         (grid_attempt_target_gridification): Convert initialization
13470         of local "grid" from memset to zero-initialization; FIXME: does
13471         this require C++11?  Update call to dump_printf_loc to pass in a
13472         optinfo_location rather than a location_t, via the gimple stmt.
13473         * profile.c (read_profile_edge_counts): Updates call to
13474         dump_printf_loc to pass in a dump_location_t rather than a
13475         location_t
13476         (compute_branch_probabilities): Likewise.
13477         * selftest-run-tests.c (selftest::run_tests): Call
13478         dumpfile_c_tests.
13479         * selftest.h (dumpfile_c_tests): New decl.
13480         * tree-loop-distribution.c (pass_loop_distribution::execute):
13481         Update for change in return type of find_loop_location.
13482         * tree-parloops.c (parallelize_loops): Likewise.
13483         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
13484         "locus" from location_t to dump_user_location_t.
13485         (canonicalize_loop_induction_variables): Likewise.
13486         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
13487         for change in return type of find_loop_location.
13488         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
13489         to dump_printf_loc to pass in a dump_location_t rather than a
13490         location_t, via the stmt.
13491         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
13492         Likewise.
13493         * tree-vect-loop-manip.c (find_loop_location): Convert return
13494         type from source_location to dump_user_location_t.
13495         (vect_do_peeling): Update for above change.
13496         (vect_loop_versioning): Update for change in type of
13497         vect_location.
13498         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
13499         from location_t to dump_user_location_t.
13500         (vect_estimate_min_profitable_iters): Update for change in type
13501         of vect_location.
13502         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
13503         location_t to dump_location_t.
13504         (vect_slp_bb): Update for change in type of vect_location.
13505         * tree-vectorizer.c (vect_location): Convert from source_location
13506         to dump_user_location_t.
13507         (try_vectorize_loop_1): Update for change in vect_location's type.
13508         (vectorize_loops): Likewise.
13509         (increase_alignment): Likewise.
13510         * tree-vectorizer.h (vect_location): Convert from source_location
13511         to dump_user_location_t.
13512         (find_loop_location): Convert return type from source_location to
13513         dump_user_location_t.
13514         (check_reduction_path): Convert 1st param from location_t to
13515         dump_user_location_t.
13516         * value-prof.c (check_counter): Update call to dump_printf_loc to
13517         pass in a dump_user_location_t rather than a location_t; update
13518         call to error_at for change in type of "locus".
13519         (check_ic_target): Update call to dump_printf_loc to
13520         pass in a dump_user_location_t rather than a location_t, via the
13521         call_stmt.
13523 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
13525         * config/s390/s390.h (enum processor_flags): Do not use
13526         default tune parameter when -march was specified.
13528 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
13530         PR target/86314
13531         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
13532         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
13533         operands.
13535 2018-06-26  Richard Biener  <rguenther@suse.de>
13537         PR tree-optimization/86287
13538         PR bootstrap/86316
13539         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
13540         (vect_analyze_loop): Initialize n_stmts.
13542 2018-06-26  Richard Biener  <rguenther@suse.de>
13544         PR middle-end/86271
13545         * fold-const.c (fold_convertible_p): Pointer extension
13546         isn't valid.
13548 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
13550         PR debug/86064
13551         * dwarf2out.c (loc_list_has_views): Adjust comments.
13552         (dw_loc_list): Split single cross-partition range with
13553         nonzero locview.
13555 2018-06-25  Jeff Law  <law@redhat.com>
13557         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
13558         on -mbig-switch by default.
13560         * config/v850/predicates.md (const_float_1_operand): Fix match_code
13561         test.
13562         (const_float_0_operand): Remove unused predicate.
13563         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
13564         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
13565         (recipsf2): New expander.  Original pattern now called
13566         (recipsf2_insn).
13567         (recipdf2, recipdf2_insn): Similarly.
13568         (rsqrtsf2, rsqrtsf2_insn): Similarly
13569         (rsqrtdf2, rsqrtdf2_insn): Similarly
13571 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
13573         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
13574         Simplify logic for FreeBSD (twice).
13576 2018-06-25  Martin Sebor  <msebor@redhat.com>
13578         PR tree-optimization/86204
13579         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
13580         a strnlen result if it's less than the length of the string.
13582 2018-06-25  Martin Sebor  <msebor@redhat.com>
13584         PR tree-optimization/85700
13585         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
13586         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
13587         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
13589 2018-06-25  Martin Sebor  <msebor@redhat.com>
13591         * doc/extend.texi (Zero-length arrays): Update and clarify.
13593 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
13595         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
13596         added IEEE/IBM long double multilib support on PowerPC little
13597         endian Linux systems.
13598         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
13599         (MULTILIB_DEFAULTS): Likewise.
13600         * config/rs6000/rs6000.c (rs6000_option_override_internal):
13601         Likewise.
13602         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
13603         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
13604         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
13606 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
13608         PR middle-end/86311
13609         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
13610         (REORDER_45): Likewise.
13612 2018-06-25  Jeff Law  <law@redhat.com>
13614         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
13615         dividend to 32 bits.  Adjust length.
13616         (udivmodhi4): Cleanup output template.  Fix length.
13618 2018-06-25  Carl Love  <cel@us.ibm.com>
13620         * config/rs6000/vsx.md: Change word selector to prefered location.
13622 2018-06-25  Richard Biener  <rguenther@suse.de>
13624         PR tree-optimization/86304
13625         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
13626         epilogue-if-converted loops as well.
13628 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
13630         * lto-section-out.c (lto_begin_section): Do not print section
13631         name for noaddr and unnumbered dumps.
13633 2018-06-25  Richard Biener  <rguenther@suse.de>
13635         * tree-vectorizer.h (struct vec_info_shared): New structure
13636         with parts split out from struct vec_info and loop_nest from
13637         struct _loop_vec_info.
13638         (struct vec_info): Adjust accordingly.
13639         (struct _loop_vec_info): Likewise.
13640         (LOOP_VINFO_LOOP_NEST): Adjust.
13641         (LOOP_VINFO_DATAREFS): Likewise.
13642         (LOOP_VINFO_DDRS): Likewise.
13643         (struct _bb_vec_info): Likewise.
13644         (BB_VINFO_DATAREFS): Likewise.
13645         (BB_VINFO_DDRS): Likewise.
13646         (struct _stmt_vec_info): Add dr_aux member.
13647         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
13648         (DR_MISALIGNMENT_UNINITIALIZED): New.
13649         (set_dr_misalignment): Adjust.
13650         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
13651         (vect_analyze_loop): Adjust prototype.
13652         (vect_analyze_loop_form): Likewise.
13653         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
13654         Compute dependences lazily.
13655         (vect_record_base_alignments): Use shared datarefs/ddrs.
13656         (vect_verify_datarefs_alignment): Likewise.
13657         (vect_analyze_data_refs_alignment): Likewise.
13658         (vect_analyze_data_ref_accesses): Likewise.
13659         (vect_analyze_data_refs): Likewise.
13660         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
13661         constructor parameter for shared part.
13662         (vect_analyze_loop_form): Pass in shared part and adjust.
13663         (vect_analyze_loop_2): Pass in storage for the number of
13664         stmts.  Move loop nest finding to the caller.  Compute
13665         datarefs lazily.
13666         (vect_analyze_loop): Pass in shared part.
13667         (vect_transform_loop): Verify shared datarefs are unchanged.
13668         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
13669         constructor parameter for shared part.
13670         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
13671         (vect_slp_bb): Verify shared datarefs are unchanged before
13672         transform.
13673         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
13674         change.
13675         (new_stmt_vec_info): Initialize DR_AUX misalignment to
13676         DR_MISALIGNMENT_UNINITIALIZED.
13677         * tree-vectorizer.c (vec_info::vec_info): Add constructor
13678         parameter for shared part.
13679         (vec_info::~vec_info): Adjust.
13680         (vec_info_shared::vec_info_shared): New.
13681         (vec_info_shared::~vec_info_shared): Likewise.
13682         (vec_info_shared::save_datarefs): Likewise.
13683         (vec_info_shared::check_datarefs): Likewise.
13684         (try_vectorize_loop_1): Construct shared part live for analyses
13685         of a single loop for multiple vector sizes.
13686         * tree-parloops.c (gather_scalar_reductions): Adjust.
13688 2018-06-25  Richard Biener  <rguenther@suse.de>
13690         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
13691         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
13692         (vect_analyze_data_refs): Remove similar code from here and
13693         simplify accordingly.
13695 2018-06-25  Richard Biener  <rguenther@suse.de>
13697         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
13698         for reverse storage order accesses rather than asserting
13699         they cannot happen here.
13701 2018-06-25  Tom de Vries  <tdevries@suse.de>
13703         PR debug/86257
13704         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
13705         Use data16 instead of .byte for insn prefix.
13707 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
13709         PR C++/86082
13710         * parser.c (make_char_string_pack): Pass this literal chars
13711         through cpp_interpret_string.
13712         (cp_parser_userdef_numeric_literal): Check the result of
13713         make_char_string_pack.
13715 2018-06-24  Maya Rashish  <coypu@sdf.org>
13717         * ginclude/stddef.h: Simplify conditions around avoiding
13718         re-definition of __size_t.
13720 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
13722         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
13723         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
13725 2018-06-22  Maya Rashish  <coypu@sdf.org>
13727         * doc/invoke.texi (mno-fancy-math-387): Update for changes
13728         made to OpenBSD and NetBSD through the years.
13730 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13732         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
13733         behavior of vec_pack (vector double, vector double) to match
13734         behavior of vec_float2 (vector double, vector double).
13736 2018-06-22  Olivier Hainque  <hainque@adacore.com>
13738         * gimplify.c (gimplify_function_tree): Prevent creation
13739         of a trampoline for the address of the current function
13740         passed to entry/exit instrumentation hooks.
13742 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
13744         PR target/86222
13745         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
13746         correctly.
13748 2018-06-22  Martin Liska  <mliska@suse.cz>
13750         PR tree-optimization/86263
13751         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
13752         Bail out if is_enabled is false.
13753         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
13754         New declaration.
13755         (jump_table_cluster::is_enabled): New function.
13757 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
13759         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
13760         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
13761         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
13762         (lto_input_ts_binfo_tree_pointers): Likewise.
13763         * tree-streamer-out.c (streamer_write_tree_bitfields,
13764         write_ts_binfo_tree_pointers): Likewise.
13765         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
13767 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
13769         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
13771 2018-06-22  Martin Liska  <mliska@suse.cz>
13773         * symbol-summary.h (get): Make it pure and inline move
13774         functionality from ::get function.
13775         (get): Remove and inline into ::get and ::get_create.
13776         (get_create): Move code from ::get function.
13778 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13780         PR target/85994
13781         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
13782         -x assembler-with-cpp.
13784 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13786         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
13787         _FILE_OFFSET_BITS=64 for C++.
13789 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
13791         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
13792         conversion insn that shows up when pr85657-3.c is compiled using
13793         IEEE 128-bit long double.
13794         (neg<mode>2_internal): Use the correct mode to check whether the
13795         mode is IBM extended.
13796         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
13797         multiply and divide external functions from being created more
13798         than once.
13800 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
13802         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
13803         functions.
13804         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
13805         the edge can be forwarded.
13806         (cfg_layout_merge_blocks): Likewise.
13808 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
13810         * except.c (finish_eh_generation): Commit edge insertions only after
13811         the EH edges have been redirected from post-landing to landing pads.
13813 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
13815         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
13816         create_tmp_var_for to create the FRAME decl.
13817         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
13819 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
13821         * tree-inline.c (copy_edges_for_bb): Minor tweak.
13822         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
13823         debug statement when resetting its value.
13824         (expand_call_inline): Copy the locus of the call onto the assignment
13825         of the return value, if any.  Use local variable in more cases.
13827 2018-06-21  Martin Liska  <mliska@suse.cz>
13829         * ipa-pure-const.c (propagate_nothrow): Use
13830         funct_state_summaries->get.
13831         (dump_malloc_lattice): Likewise.
13832         (propagate_malloc): Likewise.
13834 2018-06-21  Richard Biener  <rguenther@suse.de>
13836         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
13837         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
13838         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
13839         comment.
13840         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
13841         BLOCK_ABSTRACT_ORIGIN unconditionally.
13843 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
13845         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
13846         deleting it.
13847         * ipa-reference.c (ipa_reference_c_finalize): Delete
13848         ipa_ref_opt_sum_summaries and set it to NULL.
13850 2018-06-21  Tom de Vries  <tdevries@suse.de>
13852         PR tree-optimization/85859
13853         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
13854         test with comment from bb_no_side_effects_p.
13856 2018-06-21  Richard Biener  <rguenther@suse.de>
13858         PR tree-optimization/86232
13859         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
13860         max for constant niter.
13862 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13864         * config/aarch64/aarch64-simd.md
13865         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
13867 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13869         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
13870         Make opernads of the unspec commutative.
13872 2018-06-21  Richard Biener  <rguenther@suse.de>
13874         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
13875         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
13876         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
13877         (vect_analyze_data_ref_dependence): Re-order checks to deal with
13878         NULL DR_STEP.
13879         (vect_record_base_alignments): Do not record base alignment
13880         for gathers or scatters.
13881         (vect_compute_data_ref_alignment): Drop return value that is always
13882         true.  Bail out early for gathers or scatters.
13883         (vect_enhance_data_refs_alignment): Bail out early for gathers
13884         or scatters.
13885         (vect_find_same_alignment_drs): Likewise.
13886         (vect_analyze_data_refs_alignment): Remove dead code.
13887         (vect_slp_analyze_and_verify_node_alignment): Likewise.
13888         (vect_analyze_data_refs): For possible gathers or scatters do
13889         not create an alternate DR, just check their possible validity
13890         and mark them.  Adjust DECL_NONALIASED handling to not rely
13891         on DR_BASE_ADDRESS.
13892         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
13893         update inits of gathers or scatters.
13894         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
13895         Also copy gather/scatter flag to pattern vinfo.
13897 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13899         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
13900         behavior of vec_packsu (vector unsigned long long, vector unsigned
13901         long long) to match behavior of vec_packs with same signature.
13903 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
13904             Thomas Schwinge <thomas@codesourcery.com>
13905             Cesar Philippidis  <cesar@codesourcery.com>
13907         * gimplify.c (gimplify_scan_omp_clauses): Add support for
13908         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
13909         (gimplify_adjust_omp_clauses): Likewise.
13910         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
13911         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
13912         (gimplify_omp_target_update): Update handling of acc update and
13913         enter/exit data.
13914         * omp-low.c (install_var_field): Remove unused parameter
13915         base_pointers_restrict.
13916         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
13917         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
13918         FINALIZE}
13919         (omp_target_base_pointers_restrict_p): Delete.
13920         (scan_omp_target): Update call to scan_sharing_clauses.
13921         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
13922         FINALIZE}.
13923         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
13924         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
13925         (convert_local_omp_clauses): Likewise.
13926         * tree-pretty-print.c (dump_omp_clause): Likewise.
13927         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
13928         FINALIZE}.
13929         (omp_clause_code_name): Likewise.
13931 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
13933         PR debug/86194
13934         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
13935         be narrowed.
13937         PR tree-optimization/86231
13938         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
13939         anti-range don't overwrite *vr0min before using it to compute *vr0max.
13941 2018-06-20  Tom de Vries  <tdevries@suse.de>
13943         PR tree-optimization/86097
13944         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
13945         iv type if signedness of iv type is not the same as that of *nit.
13947 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
13949         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
13950         EDGE_EH edges, verify they are all EDGE_EH.
13952 2018-06-20  Maya Rashish  <coypu@sdf.org>
13954         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
13956 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13958         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
13959         * config/aarch64/aarch64.c (xgene1_tunings): Add
13960         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
13961         (aarch64_mode_valid_for_sched_fusion_p):
13962         Allow 16-byte modes.
13963         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
13964         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
13965         128-bit modes.
13966         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
13967         New pattern.
13968         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
13969         * config/aarch64/iterators.md (VQ2): New mode iterator.
13971 2018-06-20  Martin Liska  <mliska@suse.cz>
13973         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
13974         Change default ratio from 10 to 8.
13976 2018-06-20  Martin Liska  <mliska@suse.cz>
13978         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
13979         New.
13980         (bit_test_cluster::find_bit_tests): Likewise.
13981         (switch_decision_tree::analyze_switch_statement): Find clusters.
13982         * tree-switch-conversion.h (struct jump_table_cluster): Document
13983         hierarchy.
13985 2018-06-20  Martin Liska  <mliska@suse.cz>
13987         * tree-switch-conversion.c (switch_conversion::collect):
13988         Record m_uniq property.
13989         (switch_conversion::expand): Bail out for special conditions.
13990         (group_cluster::~group_cluster): New.
13991         (group_cluster::group_cluster): Likewise.
13992         (group_cluster::dump): Likewise.
13993         (jump_table_cluster::emit): New.
13994         (switch_decision_tree::fix_phi_operands_for_edges): New.
13995         (struct case_node): Remove struct.
13996         (jump_table_cluster::can_be_handled): New.
13997         (case_values_threshold): Moved to header.
13998         (reset_out_edges_aux): Likewise.
13999         (jump_table_cluster::is_beneficial): New.
14000         (bit_test_cluster::can_be_handled): Likewise.
14001         (add_case_node): Remove.
14002         (bit_test_cluster::is_beneficial): New.
14003         (case_bit_test::cmp): New.
14004         (bit_test_cluster::emit): New.
14005         (expand_switch_as_decision_tree_p): Remove.
14006         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
14007         (fix_phi_operands_for_edge): Likewise.
14008         (switch_decision_tree::analyze_switch_statement): New.
14009         (compute_cases_per_edge): Move ...
14010         (switch_decision_tree::compute_cases_per_edge): ... here.
14011         (try_switch_expansion): Likewise.
14012         (switch_decision_tree::try_switch_expansion): Likewise.
14013         (record_phi_operand_mapping): Likewise.
14014         (switch_decision_tree::record_phi_operand_mapping): Likewise.
14015         (emit_case_decision_tree): Likewise.
14016         (switch_decision_tree::emit): Likewise.
14017         (balance_case_nodes): Likewise.
14018         (switch_decision_tree::balance_case_nodes): Likewise.
14019         (dump_case_nodes): Likewise.
14020         (switch_decision_tree::dump_case_nodes): Likewise.
14021         (emit_jump): Likewise.
14022         (switch_decision_tree::emit_jump): Likewise.
14023         (emit_cmp_and_jump_insns): Likewise.
14024         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
14025         (emit_case_nodes): Likewise.
14026         (switch_decision_tree::emit_case_nodes): Likewise.
14027         (conditional_probability): Remove.
14028         * tree-switch-conversion.h (enum cluster_type): New.
14029         (PRINT_CASE): New.
14030         (struct cluster): Likewise.
14031         (cluster::cluster): Likewise.
14032         (struct simple_cluster): Likewise.
14033         (simple_cluster::simple_cluster): Likewise.
14034         (struct group_cluster): Likewise.
14035         (struct jump_table_cluster): Likewise.
14036         (struct bit_test_cluster): Likewise.
14037         (struct min_cluster_item): Likewise.
14038         (struct case_tree_node): Likewise.
14039         (case_tree_node::case_tree_node): Likewise.
14040         (jump_table_cluster::case_values_threshold): Likewise.
14041         (struct case_bit_test): Likewise.
14042         (struct switch_decision_tree): Likewise.
14043         (struct switch_conversion): Likewise.
14044         (switch_decision_tree::reset_out_edges_aux): Likewise.
14046 2018-06-20  Martin Liska  <mliska@suse.cz>
14048         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
14049         (hoist_edge_and_branch_if_true): Likewise.
14050         (expand_switch_using_bit_tests_p): Likewise.
14051         (struct case_bit_test): Likewise.
14052         (case_bit_test_cmp): Likewise.
14053         (emit_case_bit_tests): Likewise.
14054         (switch_conversion::switch_conversion): New class.
14055         (struct switch_conv_info): Remove old struct.
14056         (collect_switch_conv_info): More to ...
14057         (switch_conversion::collect): ... this.
14058         (check_range): Likewise.
14059         (switch_conversion::check_range): Likewise.
14060         (check_all_empty_except_final): Likewise.
14061         (switch_conversion::check_all_empty_except_final): Likewise.
14062         (check_final_bb): Likewise.
14063         (switch_conversion::check_final_bb): Likewise.
14064         (create_temp_arrays): Likewise.
14065         (switch_conversion::create_temp_arrays): Likewise.
14066         (free_temp_arrays): Likewise.
14067         (gather_default_values): Likewise.
14068         (switch_conversion::gather_default_values): Likewise.
14069         (build_constructors): Likewise.
14070         (switch_conversion::build_constructors): Likewise.
14071         (constructor_contains_same_values_p): Likewise.
14072         (switch_conversion::contains_same_values_p): Likewise.
14073         (array_value_type): Likewise.
14074         (switch_conversion::array_value_type): Likewise.
14075         (build_one_array): Likewise.
14076         (switch_conversion::build_one_array): Likewise.
14077         (build_arrays): Likewise.
14078         (switch_conversion::build_arrays): Likewise.
14079         (gen_def_assigns): Likewise.
14080         (switch_conversion::gen_def_assigns): Likewise.
14081         (prune_bbs): Likewise.
14082         (switch_conversion::prune_bbs): Likewise.
14083         (fix_phi_nodes): Likewise.
14084         (switch_conversion::fix_phi_nodes): Likewise.
14085         (gen_inbound_check): Likewise.
14086         (switch_conversion::gen_inbound_check): Likewise.
14087         (process_switch): Use the newly created class.
14088         (switch_conversion::expand): New.
14089         (switch_conversion::~switch_conversion): New.
14090         * tree-switch-conversion.h: New file.
14092 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14094         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
14095         tree-vect-patterns.c.
14096         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
14097         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
14098         (vect_recog_sad_pattern): Likewise.
14099         (vect_recog_widen_sum_pattern): Likewise.
14100         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
14101         (vect_recog_widen_shift_pattern): Remove the type_in argument.
14102         (vect_recog_rotate_pattern): Likewise.
14103         (vect_recog_mult_pattern): Likewise.
14104         (vect_recog_vector_vector_shift_pattern): Likewise.
14105         (vect_recog_divmod_pattern): Likewise.
14106         (vect_recog_mixed_size_cond_pattern): Likewise.
14107         (vect_recog_bool_pattern): Likewise.
14108         (vect_recog_mask_conversion_pattern): Likewise.
14109         (vect_try_gather_scatter_pattern): Likewise.
14110         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
14111         (vect_recog_over_widening_pattern): Likewise.
14112         (vect_recog_gather_scatter_pattern): Likewise.
14113         (vect_recog_func_ptr): Move from tree-vectorizer.h
14114         (vect_vect_recog_func_ptrs): Move further down the file.
14115         (vect_recog_func): Likewise.  Remove the third argument.
14116         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
14117         (vect_pattern_recog_1): Expect the pattern function to do any
14118         necessary target tests.  Also expect it to provide a vector type.
14119         Remove the type_in handling.
14121 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14123         * tree-vect-patterns.c (vect_pattern_detected): New function.
14124         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
14125         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
14126         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
14127         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
14128         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
14129         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
14130         (vect_recog_mask_conversion_pattern)
14131         (vect_try_gather_scatter_pattern): Likewise.
14133 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14135         * tree-vect-patterns.c (vect_get_internal_def): New function.
14136         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14137         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
14138         (search_type_for_mask_1): Use it.
14140 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14142         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
14143         redundant WIDEN_SUM_EXPR handling.
14144         (vect_recog_sad_pattern): Likewise.
14146 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14148         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
14149         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
14150         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
14151         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
14152         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
14154 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14156         * tree-vect-stmts.c (vectorizable_call): Make sure that we
14157         use the stmt_vec_info of the original bb statement for the
14158         new zero assignment, even if the call is part of a pattern.
14160 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14162         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
14163         that the sequence is attached to the original statement rather
14164         than the pattern statement.
14165         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
14166         PATTERN_DEF_SEQ from the original statement rather than
14167         the main pattern statement.
14168         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
14169         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
14170         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
14172 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
14174         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
14175         definition statements before the early exit for statements that aren't
14176         live or relevant.
14177         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
14178         split out from...
14179         (vect_transform_loop): ...here.  Process pattern definition
14180         statements without first checking whether the main pattern
14181         statement is live or relevant.
14183 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
14185         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
14186         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
14188 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
14190         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
14191         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
14192         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
14193         (expand_block_compare): Change select_block_compare_mode call.
14194         (expand_strncmp_align_check): Use new functions, fix comment.
14195         (emit_final_str_compare_gpr): New function.
14196         (expand_strn_compare): Refactor and clean up code.
14197         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
14199 2018-06-19  Tony Reix  <tony.reix@atos.com>
14200             Damien Bergamini  <damien.bergamini@atos.com>
14201             David Edelsohn  <dje.gcc@gmail.com>
14203         * collect2.c (static_obj): New variable.
14204         (static_libs): New variable.
14205         (is_in_list): Uncomment declaration.
14206         (main): Track AIX libraries linked statically.
14207         (is_in_list): Uncomment definition.
14208         (scan_prog_file): Don't add AIX shared libraries initializer
14209         to constructor list if linking statically.
14211 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
14213         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
14214         constant.
14215         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
14217 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
14219         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
14220         blocks.
14222 2018-06-19  Martin Liska  <mliska@suse.cz>
14224         * config/i386/i386.c (ix86_can_inline_p): Do not use
14225         ipa_fn_summaries::get_create.
14226         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
14227         get.
14228         (devirtualization_time_bonus): Likewise.
14229         (ipcp_propagate_stage): Likewise.
14230         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
14231         (edge_set_predicate): Likewise.
14232         (evaluate_conditions_for_known_args): Likewise.
14233         (evaluate_properties_for_edge): Likewise.
14234         (ipa_call_summary::reset): Tranform to ...
14235         (ipa_call_summary::~ipa_call_summary): ... this.
14236         (ipa_fn_summary::reset): Transform to ...
14237         (ipa_fn_summary::~ipa_fn_summary): ... this.
14238         (ipa_fn_summary_t::remove): Rename to ...
14239         (ipa_fn_summary_t::remove_callees): ... this.
14240         (ipa_fn_summary_t::duplicate): Use placement new
14241         instead of memory copy.
14242         (ipa_call_summary_t::duplicate): Likewise.
14243         (ipa_call_summary_t::remove): Remove.
14244         (dump_ipa_call_summary): Change get_create to get.
14245         (ipa_dump_fn_summary): Dump only when summary exists.
14246         (analyze_function_body): Use symbol_summary::get instead
14247         of get_create.
14248         (compute_fn_summary): Likewise.
14249         (estimate_edge_devirt_benefit): Likewise.
14250         (estimate_edge_size_and_time): Likewise.
14251         (inline_update_callee_summaries): Likewise.
14252         (remap_edge_change_prob): Likewise.
14253         (remap_edge_summaries): Likewise.
14254         (ipa_merge_fn_summary_after_inlining): Likewise.
14255         (write_ipa_call_summary): Likewise.
14256         (ipa_fn_summary_write): Likewise.
14257         (ipa_free_fn_summary): Likewise.
14258         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
14259         (struct ipa_call_summary): Likewise.
14260         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
14261         of get_create.
14262         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
14263         (estimate_size_after_inlining): Likewise.
14264         (estimate_growth): Likewise.
14265         (growth_likely_positive): Likewise.
14266         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
14267         (inline_call): Likewise.
14268         * ipa-inline.c (caller_growth_limits): Likewise.
14269         (can_inline_edge_p): Likewise.
14270         (can_inline_edge_by_limits_p): Likewise.
14271         (compute_uninlined_call_time): Likewise.
14272         (compute_inlined_call_time): Likewise.
14273         (want_inline_small_function_p): Likewise.
14274         (edge_badness): Likewise.
14275         (update_caller_keys): Likewise.
14276         (update_callee_keys): Likewise.
14277         (inline_small_functions): Likewise.
14278         (inline_to_all_callers_1): Likewise.
14279         (dump_overall_stats): Likewise.
14280         (early_inline_small_functions): Likewise.
14281         (early_inliner): Likewise.
14282         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
14283         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14284         * ipa-pure-const.c (malloc_candidate_p): Likewise.
14285         * ipa-split.c (execute_split_functions): Likewise.
14286         * symbol-summary.h: Likewise.
14287         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
14289 2018-06-19  Richard Biener  <rguenther@suse.de>
14291         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
14292         (vectorize_loops): ... here.  Fix dbgcnt handling.
14293         (try_vectorize_loop): Wrap try_vectorize_loop_1.
14295 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
14297         PR target/86197
14298         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
14299         ieee128 argument takes up only one (vector) register, not two (floating
14300         point) registers.
14302 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
14304         * gimplify.c (gimplify_init_constructor): Really never clear for an
14305         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
14307 2018-06-19  Richard Biener  <rguenther@suse.de>
14309         PR tree-optimization/86179
14310         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
14311         after failed recognition.
14313 2018-06-18  Martin Sebor  <msebor@redhat.com>
14315         PR middle-end/85602
14316         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
14317         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
14318         Handle integer subtraction.
14319         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
14320         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
14322 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
14324         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
14325         param from rtx to rtx_insn *.
14326         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
14327         param.
14328         (frv_ifcvt_modify_insn): Likwise.
14329         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
14330         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
14331         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
14332         as_a <rtx_insn *> cast to local "unprotected_region" once
14333         it's been established that it's not NULL or pc_rtx.
14334         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
14335         param "sethi" from rtx to rtx_insn *.
14336         (nds32_group_float_insns): Likewise for param "insn".
14337         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
14338         param.
14339         (vax_output_int_subtract): Likewise.
14340         * config/vax/vax.c (vax_output_int_add): Likewise for param
14341         "insn".
14342         (vax_output_int_subtract): Likewise.
14343         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
14344         (emit_pattern_after): Likewise for param "after".
14345         (emit_insn_after): Likewise.
14346         (emit_jump_insn_after): Likewise.
14347         (emit_call_insn_after): Likewise.
14348         (emit_debug_insn_after): Likewise.
14349         (emit_pattern_before): Likewise for param "before".
14350         (emit_insn_before): Likewise.
14351         (emit_jump_insn_before): Likewise.
14352         * final.c (get_insn_template): Likewise for param "insn", removing
14353         a cast.
14354         * output.h (get_insn_template): Likewise for 2nd param.
14355         * rtl.h (emit_insn_before): Likewise.
14356         (emit_jump_insn_before): Likewise.
14357         (emit_debug_insn_before_noloc): Likewise.
14358         (emit_insn_after): Likewise.
14359         (emit_jump_insn_after): Likewise.
14360         (emit_call_insn_after): Likewise.
14361         (emit_debug_insn_after): Likewise.
14362         (set_insn_deleted): Likewise for param.
14364 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
14366         PR target/85358
14367         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
14368         floating point modes, so that IFmode is numerically greater than
14369         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
14370         to declare the ordering.  This prevents IFmode from being
14371         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
14372         machine.  Include rs6000-modes.h to share the fractional values
14373         between genmodes* and the rest of the compiler.
14374         (IFmode): Likewise.
14375         (KFmode): Likewise.
14376         (TFmode): Likewise.
14377         * config/rs6000/rs6000-modes.h: New file.
14378         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
14379         meaning of rs6000_long_double_size so that 126..128 selects an
14380         appropriate 128-bit floating point type.
14381         (rs6000_option_override_internal): Likewise.
14382         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
14383         (TARGET_LONG_DOUBLE_128): Change the meaning of
14384         rs6000_long_double_size so that 126..128 selects an appropriate
14385         128-bit floating point type.
14386         (LONG_DOUBLE_TYPE_SIZE): Update comment.
14387         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
14388         source and destination to match the standard usage.
14389         (truncifkf2): Likewise.
14390         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
14391         ISA 2.07 to use an explicit clobber, instead of passing in a
14392         temporary.
14393         (copysign<mode>3_soft): Likewise.
14395 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
14397         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
14398         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
14399         (vect_slp_analyze_instance_dependence): Likewise.
14400         (vect_enhance_data_refs_alignment): Likewise.
14401         (vect_analyze_data_refs_alignment): Likewise.
14402         (vect_slp_analyze_and_verify_instance_alignment
14403         (vect_analyze_data_ref_accesses): Likewise.
14404         (vect_prune_runtime_alias_test_list): Likewise.
14405         (vect_analyze_data_refs): Likewise.
14406         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
14407         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14408         (vect_analyze_scalar_cycles_1): Likewise.
14409         (vect_get_loop_niters): Likewise.
14410         (vect_analyze_loop_form_1): Likewise.
14411         (vect_update_vf_for_slp): Likewise.
14412         (vect_analyze_loop_operations): Likewise.
14413         (vect_analyze_loop): Likewise.
14414         (vectorizable_induction): Likewise.
14415         (vect_transform_loop): Likewise.
14416         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
14417         * tree-vect-slp.c (vect_analyze_slp): Likewise.
14418         (vect_make_slp_decision): Likewise.
14419         (vect_detect_hybrid_slp): Likewise.
14420         (vect_slp_analyze_operations): Likewise.
14421         (vect_slp_bb): Likewise.
14422         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
14423         (vectorizable_bswap): Likewise.
14424         (vectorizable_call): Likewise.
14425         (vectorizable_simd_clone_call): Likewise.
14426         (vectorizable_conversion): Likewise.
14427         (vectorizable_assignment): Likewise.
14428         (vectorizable_shift): Likewise.
14429         (vectorizable_operation): Likewise.
14430         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
14432 2018-06-18  Martin Sebor  <msebor@redhat.com>
14434         PR tree-optimization/81384
14435         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
14436         * builtins.c (expand_builtin_strnlen): New function.
14437         (expand_builtin): Call it.
14438         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
14439         * builtins.def (BUILT_IN_STRNLEN): New.
14440         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
14441         Warn for bounds in excess of maximum object size.
14442         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
14443         single-value ranges.  Handle strnlen.
14444         (handle_builtin_strlen): Handle strnlen.
14445         (strlen_check_and_optimize_stmt): Same.
14446         * doc/extend.texi (Other Builtins): Document strnlen.
14448 2018-06-18  Maya Rashish  <coypu@sdf.org>
14450         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
14451         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
14452         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
14454         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
14455         here to ...
14456         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
14458 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14460         * tree.c (escaped_string::escape): Replace cast to char * by
14461         const_cast<char *> (unescaped).
14463 2018-06-18  Nick Clifton  <nickc@redhat.com>
14465         PR 84195
14466         * tree.c (escaped_string): New class.  Converts an unescaped
14467         string into its escaped equivalent.
14468         (warn_deprecated_use): Use the new class to convert the
14469         deprecation message, if present.
14470         (test_escaped_strings): New self test.
14471         (test_c_tests): Add test_escaped_strings.
14472         * doc/extend.texi (deprecated): Add a note that the
14473         deprecation message is affected by the -fmessage-length
14474         option, and that control characters will be escaped.
14475         (#pragma GCC error): Document this pragma.
14476         (#pragma GCC warning): Likewise.
14477         * doc/invoke.texi (-fmessage-length): Document this option's
14478         effect on the #warning and #error preprocessor directives and
14479         the deprecated attribute.
14481 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
14483         * tree.c (decl_value_expr_lookup): Revert latest change.
14484         (decl_value_expr_insert): Likewise.
14486 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
14488         * gimplify.c (nonlocal_vlas): Delete.
14489         (nonlocal_vla_vars): Likewise.
14490         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
14491         referenced VLAs.
14492         (gimplify_body): Do not create and destroy nonlocal_vlas.
14493         * tree-nested.c: Include diagnostic.h.
14494         (use_pointer_in_frame): Tweak.
14495         (lookup_field_for_decl): Add assertion and declare the transformation.
14496         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
14497         internal error when the reference is in a wrong context.  Do not
14498         create a debug decl by default.
14499         (note_nonlocal_block_vlas): Delete.
14500         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
14501         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
14502         create a debug decl by default.
14503         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
14504         call is in a wrong context.
14505         (fixup_vla_decls): New function.
14506         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
14507         debug variables were created.
14508         * tree.c (decl_value_expr_lookup): Add checking assertion.
14509         (decl_value_expr_insert): Likewise.
14511 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
14513         PR middle-end/82479
14514         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
14515         * tree-scalar-evolution.c (interpret_expr): Likewise.
14516         (expression_expensive_p): Likewise.
14517         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
14518         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
14519         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
14520         (ssa_defined_by_minus_one_stmt_p): New.
14522 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
14524         PR middle-end/64946
14525         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
14526         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
14527         * dojump.c (do_jump): Likewise.
14528         * expr.c (expand_expr_real_2): Check operand type's sign.
14529         * fold-const.c (const_unop): Handle ABSU_EXPR.
14530         (fold_abs_const): Likewise.
14531         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
14532         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
14533         (strip_sign_op_1): Likesise.
14534         * match.pd: Add new pattern to generate ABSU_EXPR.
14535         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
14536         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
14537         * tree-eh.c (operation_could_trap_helper_p): Likewise.
14538         * tree-inline.c (estimate_operator_cost): Likewise.
14539         * tree-pretty-print.c (dump_generic_node): Likewise.
14540         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
14541         * tree.def (ABSU_EXPR): New.
14543 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
14545         PR middle-end/86095
14546         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
14547         documented as preserved for backward compatibility only.
14548         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
14550         PR rtl-optimization/86108
14551         * bb-reorder.c (create_forwarder_block): Renamed to ...
14552         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
14553         jump from new landing pad to the second part.
14554         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
14555         Adjust callers.
14557 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
14559         PR middle-end/85878
14560         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
14561         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
14562         Only call store_expr for halves if the mode is the same.
14564         PR middle-end/86123
14565         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
14566         Fix up comment formatting.
14568 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14570         * typed-splay-tree.h (typed_splay_tree::remove): New function.
14571         (typed_splay_tree::closure,
14572         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
14573         (typed_splay_tree::typed_splay_tree,
14574         typed_splay_tree::operator =): Declared private.
14575         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
14576         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
14577         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
14578         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
14579         typed_splay_tree::splay_tree_splay,
14580         typed_splay_tree::splay_tree_foreach_helper,
14581         typed_splay_tree::splay_tree_insert,
14582         typed_splay_tree::splay_tree_remove,
14583         typed_splay_tree::splay_tree_lookup,
14584         typed_splay_tree::splay_tree_predecessor,
14585         typed_splay_tree::splay_tree_successor,
14586         typed_splay_tree::splay_tree_min,
14587         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
14588         (typed_splay_tree::root, typed_splay_tree::comp,
14589         typed_splay_tree::delete_key,
14590         typed_splay_tree::delete_value): New data members.
14591         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
14592         typed_splay_tree::remove.
14594 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
14596         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
14597         -mginv and -mno-ginv to the assembler.
14598         * config/mips/mips.opt (-mcrc): New option.
14599         (-mginv): Likewise.
14600         * doc/invoke.text (-mcrc): Document.
14601         (-mginv): Likewise.
14603 2018-06-15  Nick Clifton  <nickc@redhat.com>
14605         PR 84195
14606         * tree.c (escaped_string): New class.  Converts an unescaped
14607         string into its escaped equivalent.
14608         (warn_deprecated_use): Use the new class to convert the
14609         deprecation message, if present.
14610         (test_escaped_strings): New self test.
14611         (test_c_tests): Add test_escaped_strings.
14612         * doc/extend.texi (deprecated): Add a note that the
14613         deprecation message is affected by the -fmessage-length
14614         option, and that control characters will be escaped.
14615         (#pragma GCC error): Document this pragma.
14616         (#pragma GCC warning): Likewise.
14617         * doc/invoke.texi (-fmessage-length): Document this option's
14618         effect on the #warning and #error preprocessor directives and
14619         the deprecated attribute.
14621 2018-06-15  Richard Biener  <rguenther@suse.de>
14623         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
14624         here, also noting vector size used.
14625         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
14626         size used in MSG_OPTIMIZED_LOCATIONS dump.
14627         (pass_slp_vectorize::execute): Adjust.
14629 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
14631         PR target/85968
14632         * config/arc/arc.c (arc_return_address_register): Fix
14633         if-condition.
14635 2018-06-15  Richard Biener  <rguenther@suse.de>
14637         PR middle-end/86159
14638         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
14639         leave useless conversion stripping to force_gimple_operand_gsi.
14640         (gimplify_build2): Likewise.
14641         (gimplify_build1): Likewise.
14643 2018-06-15  Richard Biener  <rguenther@suse.de>
14645         PR middle-end/86076
14646         * tree-cfg.c (move_stmt_op): unshare invariant addresses
14647         before adjusting their block.
14649 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
14651         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
14652         multilibs for *-*-rtems*.
14653         * config/riscv/t-rtems: New file.
14655 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
14657         PR middle-end/86122
14658         * match.pd ((A +- CST1) +- CST2): Punt if last resort
14659         unsigned_type_for returns NULL.
14661         PR target/85945
14662         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
14663         subregs of multi-word pseudos unless the float mode has word size.
14665 2018-06-14  Richard Biener  <rguenther@suse.de>
14667         PR middle-end/86139
14668         * tree-vect-generic.c (build_word_mode_vector_type): Remove
14669         duplicate and harmful type_hash_canon.
14670         * tree.c (type_hash_canon): Assert we didn't find ourselves.
14672 2018-06-14  Richard Biener  <rguenther@suse.de>
14674         PR ipa/86124
14675         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
14676         NULL cgraph_node.
14678 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
14680         * config/rtems.h (STDINT_LONG32): Define.
14682 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
14683             Prachi Godbole  <prachi.godbole@imgtec.com>
14685         * config/mips/mips-cpus.def: Define P6600.
14686         * config/mips/mips-tables.opt: Regenerate.
14687         * config/mips/mips.c (mips_ucbranch_type): New enum.
14688         (mips_rtx_cost_data): Add support for P6600.
14689         (mips_issue_rate): Likewise.
14690         (mips_multipass_dfa_lookahead): Likewise.
14691         (mips_avoid_hazard): Likewise.
14692         (mips_reorg_process_insns): Likewise.
14693         (mips_classify_branch_p6600): New function.
14694         * config/mips/mips.h (TUNE_P6600): New define.
14695         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
14696         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
14697         * config/mips/mips.md: Include p6600.md.
14698         (processor): Add p6600.
14699         * config/mips/p6600.md: New file.
14700         * doc/invoke.texi: Add p6600 to supported architectures.
14702 2018-06-13  Martin Sebor  <msebor@redhat.com>
14704         PR tree-optimization/86114
14705         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
14706         of integer types.
14707         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
14709 2018-06-13  Richard Biener  <rguenther@suse.de>
14711         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
14712         Properly set vector type of the intermediate stmt.
14713         * tree-vect-stmts.c (vectorizable_operation): The destination
14714         var always has vectype_out type.
14716 2018-06-13  Jeff Law  <law@redhat.com>
14718         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
14719         integer 0 for argument to print_rtl_with_bb.
14720         (rl78_reorg): Likewise.
14722 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
14724         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
14725         from rtx to rtx_insn *.
14726         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
14727         "label".
14728         (add_sched_insns_for_speculation): Likewise for local "target",
14729         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
14730         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
14731         from rtx_insn ** to rtx_code_label **.
14732         (reorg_emit_nops): Likewise.
14733         (c6x_reorg): Likewise for local "call_labels".
14734         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
14735         rtx to rtx_insn *.
14736         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
14737         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
14738         the loops over LABEL_REFS.
14739         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
14740         braf_label.
14741         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
14742         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
14743         (split_branches): Strengthen local "olabel" from rtx to
14744         rtx_insn *, adding a safe_as_a cast.
14745         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
14746         to "rtx_insn *".
14747         (add_insn_after): Likewise for first two params.
14748         (add_insn_before): Likewise.
14749         (remove_insn): Likewise for param.
14750         (emit_pattern_before_noloc): Likewise for second and third params.
14751         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
14752         (emit_call_insn_before_noloc): Likewise.
14753         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
14754         to "rtx_insn *".
14755         (emit_barrier_before): Likewise.
14756         (emit_label_before): Strengthen "label" param from "rtx" to
14757         "rtx_code_label *".  Strengthen "before" param from "rtx" to
14758         "rtx_insn *".
14759         (emit_insn_after_1): Strengthen "after" param from "rtx" to
14760         "rtx_insn *".
14761         (emit_pattern_after_noloc): Likewise.
14762         (emit_insn_after_noloc): Likewise.
14763         (emit_jump_insn_after_noloc): Likewise.
14764         (emit_call_insn_after_noloc): Likewise.
14765         (emit_debug_insn_after_noloc): Likewise.
14766         (emit_barrier_after): Likewise.
14767         (emit_label_after): Likewise for both params.
14768         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
14769         "loc" param from "int" to "location_t".
14770         (emit_insn_after_setloc): Likewise.
14771         (emit_jump_insn_after_setloc): Likewise.
14772         (emit_call_insn_after_setloc): Likewise.
14773         (emit_debug_insn_after_setloc): Likewise.
14774         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
14775         "loc" param from "int" to "location_t".
14776         (emit_pattern_before): Convert NULL_RTX to NULL.
14777         (emit_insn_before_setloc): Convert "loc" param from "int" to
14778         "location_t".
14779         (emit_jump_insn_before_setloc): Likewise.
14780         (emit_call_insn_before_setloc): Likewise.
14781         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
14782         rtx_insn *.  Convert "loc" param from "int" to "location_t".
14783         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
14784         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
14785         Convert 3rd param from "int" to "location_t".
14786         (emit_barrier_before, emit_barrier_after, next_real_insn):
14787         Strengthen param from rtx to rtx_insn *.
14788         (emit_label_before): Strengthen 1st param from "rtx" to
14789         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
14790         "rtx_insn *".
14791         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
14792         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
14793         Strengthen 2nd param from "rtx" to "rtx_insn *".
14794         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
14795         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
14796         Likewise. Convert 3rd param from "int" to "location_t".
14797         (emit_label_after): Strengthen 1st param from "rtx" to
14798         "rtx_code_label *".
14799         (next_real_insn, remove_insn): Strengthen param from "rtx" to
14800         "rtx_insn *".
14801         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
14802         from "rtx" to "rtx_insn *".
14804 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
14806         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
14807         bodies streamed in with -Q.
14808         * dumpfile.c (dump_files): Add lto-stream-out dump file.
14809         * dumpfile.h (tree_dump_index): Add lto_stream_out.
14810         * gimple-streamer-out.c: Include gimple-pretty-print.h
14811         (output_bb): Dump stmts streamed.
14812         * lto-section-out.c: Include print-tree.h
14813         (lto_begin_section): Dump sections created.
14814         (lto_output_decl_index): Dump decl encoded.
14815         * lto-streamer-out.c: Include print-tree.h
14816         (create_output_block): Dump output block created.
14817         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
14818         (output_function): Dump function output.
14819         (output_constructor): Dump constructor streamed.
14820         (write_global_stream): Output indexes encoded.
14821         (produce_asm_for_decls): Dump streams encoded.
14822         * lto-streamer.c (streamer_dump_file): New global var.
14823         * lto-streamer.h (streamer_dump_file): Declare.
14824         * passes.c (ipa_write_summaries): Initialize streamer dump.
14825         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
14826         in.
14828 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
14830         PR target/86048
14831         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
14832         offsets for register save directives.  Emit a second batch of save
14833         directives, if need be, when the function accesses prior frames.
14835 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
14837         * config/arc/fpu.md (fmasf4): Force operand to register.
14838         (fnmasf4): Likewise.
14840 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
14842         * config/arc/arc-protos.h (arc_pad_return): Remove.
14843         * config/arc/arc.c (machine_function): Remove force_short_suffix
14844         and size_reason.
14845         (arc_print_operand): Adjust printing of '&'.
14846         (arc_verify_short): Remove conditional printing of short suffix.
14847         (arc_final_prescan_insn): Remove reference to size_reason.
14848         (pad_return): New function.
14849         (arc_reorg): Call pad_return.
14850         (arc_pad_return): Remove.
14851         (arc_init_machine_status): Remove reference to force_short_suffix.
14852         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
14853         (attr length): When attribute iscompact is true force to 2
14854         regardless; in the case of maybe check if we want to force the
14855         instruction to have 4 bytes length.
14856         (nopv): Change it to generate 4 byte long nop as well.
14857         (blockage): New pattern.
14858         (simple_return): Remove call to arc_pad_return.
14859         (p_return_i): Likewise.
14861 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
14863         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
14865 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
14867         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
14868         ARC cores.
14870 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
14872         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
14873         for ARC700 and ARCv2.
14875 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
14877         PR target/86076
14878         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
14879         operands[2] instead of operands[1].
14882 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
14884         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
14885         case, check whether the outer register overlaps an unallocatable
14886         register, not just whether it fits the required class.
14888 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
14890         * poly-int.h (can_div_trunc_p): Add new overload in which all values
14891         are poly_ints.
14892         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
14893         (memrefs_conflict_p): Likewise.
14894         (init_alias_analysis): Likewise.
14895         * cfgexpand.c (expand_debug_expr): Likewise.
14896         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
14897         * cse.c (fold_rtx): Likewise.
14898         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
14899         * expr.c (emit_block_move_hints): Likewise.
14900         (clear_storage_hints, push_block, emit_push_insn): Likewise.
14901         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
14902         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
14903         (emit_group_store): Likewise.
14904         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
14905         to read the PRE/POST_MODIFY increment.
14906         * calls.c (store_one_arg): Use strip_offset.
14907         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
14908         poly_int_rtx_p.
14909         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
14910         by a VEC_SELECT.
14911         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
14912         (simplify_binary_operation_1): Extend CONST_INT handling to
14913         poly_int_rtx_p.
14914         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
14915         than a HOST_WIDE_INT.
14916         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
14917         poly_int64.
14918         (adjust_mems, add_stores): Update accodingly.
14919         (vt_canonicalize_addr): Track polynomial offsets.
14920         (emit_note_insn_var_location): Likewise.
14921         (vt_add_function_parameter): Likewise.
14922         (vt_initialize): Likewise.
14924 2018-06-12  Jeff Law  <law@redhat.com>
14926         * config.gcc (alpha*-*-freebsd*): Remove.
14927         * config/alpha/freebsd.h: Remove.
14929 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
14931         PR other/69968
14932         * spellcheck-tree.c (levenshtein_distance): Rename to...
14933         (get_edit_distance): ...this, and update for underlying renaming.
14934         * spellcheck-tree.h (levenshtein_distance): Rename to...
14935         (get_edit_distance): ...this.
14936         * spellcheck.c (levenshtein_distance): Rename to...
14937         (get_edit_distance): ...this.  Convert from Levenshtein distance
14938         to Damerau-Levenshtein distance by supporting transpositions of
14939         adjacent characters.  Rename "v1" to "v_next" and "v0" to
14940         "v_one_ago".
14941         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
14942         (selftest::test_edit_distance_unit_test_oneway): ...this, and
14943         update for underlying renaming.
14944         (selftest::levenshtein_distance_unit_test): Rename to...
14945         (selftest::test_get_edit_distance_unit): ...this, and update for
14946         underlying renaming.
14947         (selftest::test_find_closest_string): Add example from PR 69968
14948         where transposition helps
14949         (selftest::test_metric_conditions): Update for renaming.
14950         (selftest::test_metric_conditions): Likewise.
14951         (selftest::spellcheck_c_tests): Likewise.
14952         * spellcheck.h (levenshtein_distance): Rename both overloads to...
14953         (get_edit_distance): ...this.
14954         (best_match::consider): Update for renaming.
14956 2018-06-12  Martin Sebor  <msebor@redhat.com>
14958         PR tree-optimization/85259
14959         * builtins.c (compute_objsize): Handle constant offsets.
14960         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
14961         true iff a warning has been issued.
14962         * gimple.h (gimple_nonartificial_location): New function.
14963         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
14964         gimple_nonartificial_location and handle -Wno-system-headers.
14965         (handle_builtin_stxncpy): Same.
14967 2018-06-12  Martin Sebor  <msebor@redhat.com>
14969         PR c/85931
14970         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
14972 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
14974         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
14975         BUILTIN_VEC_XST entries for pointer to double and long long.
14977 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
14979         PR target/85990
14980         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
14981         Update comments.
14982         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
14983         Likewise.
14985 2018-06-12  Martin Liska  <mliska@suse.cz>
14987         * doc/options.texi: Document IntegerRange.
14989 2018-06-12  Martin Liska  <mliska@suse.cz>
14991         * config/i386/i386.opt: Make MPX-related options as Deprecated.
14992         * opt-functions.awk: Handle Deprecated flag.
14993         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
14994         and report error.
14995         (read_cmdline_option): Report warning for a deprecated option.
14996         * opts.h (struct cl_option): Add new field cl_deprecated.
14997         (CL_ERR_DEPRECATED): New.
14999 2018-06-12  Martin Liska  <mliska@suse.cz>
15001         * doc/options.texi: Document Deprecated option flag.
15003 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15005         * config/arc/arc-arch.h (arc_extras): New enum.
15006         (arc_cpu_t):Add field extra.
15007         (arc_cpu_types): Consider the extras.
15008         * config/arc/arc-cpus.def: Add extras info.
15009         * config/arc/arc-opts.h (processor_type): Consider extra field.
15010         * config/arc/arc.c (arc_override_options): Handle extra field.
15012 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
15014         * config/arc/arc-arch.h: Update ARC_OPTX macro.
15015         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
15016         field.
15017         * config/arc/arc.c (arc_init): Update pic warning.
15018         (irq_range): Update irq range parsing warnings.
15019         (arc_override_options): Update various warning messages.
15020         (arc_handle_aux_attribute): Likewise.
15022 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
15024         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
15026 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15028         * doc/sourcebuild.texi: Document usage of line number 0 in verify
15029         compiler messages directives.
15031 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
15033         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
15034         * config/mips/mips-tables.opt: Regenerate.
15035         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
15036         mips64r6.
15037         * doc/invoke.texi: Document -march=i6500.
15039 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
15041         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
15042         (i6400_gpmul): Add cpu_unit.
15043         (i6400_gpdiv): Likewise.
15044         (i6400_msa_add_d): Update reservations.
15045         (i6400_msa_int_add) Likewise.
15046         (i6400_msa_short_logic3) Likewise.
15047         (i6400_msa_short_logic2) Likewise.
15048         (i6400_msa_short_logic) Likewise.
15049         (i6400_msa_move) Likewise.
15050         (i6400_msa_cmp) Likewise.
15051         (i6400_msa_short_float2) Likewise.
15052         (i6400_msa_div_d) Likewise.
15053         (i6400_msa_long_logic1) Likewise.
15054         (i6400_msa_long_logic2) Likewise.
15055         (i6400_msa_mult) Likewise.
15056         (i6400_msa_long_float2) Likewise.
15057         (i6400_msa_long_float4) Likewise.
15058         (i6400_msa_long_float5) Likewise.
15059         (i6400_msa_long_float8) Likewise.
15060         (i6400_fpu_fadd): Include frint type.
15061         (i6400_fpu_store): New define_insn_reservation.
15062         (i6400_fpu_load): Likewise.
15063         (i6400_fpu_move): Likewise.
15064         (i6400_fpu_fcmp): Likewise.
15065         (i6400_fpu_fmadd): Likewise.
15066         (i6400_int_mult): Include imul3nc type and update reservation.
15067         (i6400_int_div): Include idiv3 type and update reservation.
15068         (i6400_int_load): Update to check type not move_type.
15069         (i6400_int_store): Likewise.
15070         (i6400_int_prefetch): Set zero latency.
15072 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
15074         * gcc.c: Document new %@{...} sequence.
15075         (LINK_COMMAND_SPEC): Use it for the -L switches.
15076         (cpp_unique_options): Use it for the -I switches.
15077         (at_file_argbuf): New global variable.
15078         (in_at_file): Likewise.
15079         (alloc_args): Create at_file_argbuf.
15080         (clear_args): Truncate at_file_argbuf.
15081         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
15082         (open_at_file): New function.
15083         (close_at_file): Likewise.
15084         (create_at_file): Delete.
15085         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
15086         <'o'>: Likewise.
15087         <'@'>: New case.
15088         (validate_switches_from_spec): Deal with %@{...} sequence.
15089         (validate_switches): Likewise.
15090         (driver::finalize): Call clear_args.
15092 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
15094         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
15096 2018-06-11  Martin Sebor  <msebor@redhat.com>
15098         * doc/invoke.texi (-Wall): List -Wc++17-compat.
15099         (Wno-class-memaccess): Add @opindex.
15100         (Wno-templates, Wno-multiple-inheritance): Same.
15101         (Wno-virtual-inheritance, Wno-namespaces): Same.
15102         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
15103         (Wno-format-overflow, Wno-format-truncation): Same.
15104         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
15105         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
15106         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
15107         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
15108         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
15109         (Wno-misspelled-isr): Same.
15111 2018-06-11  Martin Sebor  <msebor@redhat.com>
15113         * PR tree-optimization/86083
15114         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
15116 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
15118         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
15120 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
15122         PR target/85755
15123         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
15124         on the correct operand.
15125         (*movdi_internal64): Ditto.
15127 2018-06-11  Martin Liska  <mliska@suse.cz>
15129         PR tree-optimization/86089
15130         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
15132 2018-06-11  Julia Koval  <julia.koval@intel.com>
15134         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
15135         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
15136         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
15138 2018-06-11  Olivier Hainque  <hainque@adacore.com>
15140         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
15141         for Ada with strict dwarf2.
15143 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
15145         PR target/85755
15146         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
15147         addresses.
15149 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
15151         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
15153 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
15155         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
15156         TARGET_ELF.
15158 2018-06-08  Martin Liska  <mliska@suse.cz>
15160         * tree-cfg.h (debug_function): Fix argument type to match
15161         implementation.
15163 2018-06-08  Martin Liska  <mliska@suse.cz>
15165         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
15166         Remove usage of MPX-related (and removed) fields.
15167         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
15169 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
15171         * cfg.c (debug): Use TDF_NONE rather than 0.
15172         * cfghooks.c (debug): Likewise.
15173         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
15174         (struct dump_option_value_info): Convert to...
15175         (struct kv_pair): ...this template type.
15176         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
15177         rather than 0.
15178         (optinfo_verbosity_options): Likewise.
15179         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
15180         OPTGROUP_NONE.
15181         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
15182         than int for "optgroup_flags" param.
15183         (dump_generic_expr_loc): Use dump_flags_t rather than int for
15184         "dump_kind" param.
15185         (dump_dec): Likewise.
15186         (dump_finish): Use TDF_NONE rather than 0.
15187         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
15188         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
15189         than 0.  Update for change to option_ptr.
15190         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
15191         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
15192         0.  Update for changes to optinfo_verbosity_options and
15193         optgroup_options.
15194         (opt_info_switch_p): Convert optgroup_flags from int to
15195         optgroup_flags_t.
15196         (dump_basic_block): Use dump_flags_t rather than int
15197         for "dump_kind" param.
15198         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
15199         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
15200         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
15201         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
15202         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
15203         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
15204         TDF_NONE): Convert from macros to...
15205         (enum dump_flag): ...this new enum.
15206         (dump_flags_t): Update to use enum.
15207         (operator|, operator&, operator~, operator|=, operator&=):
15208         Implement for dump_flags_t.
15209         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
15210         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
15211         Convert from macros to...
15212         (enum optgroup_flag): ...this new enum.
15213         (optgroup_flags_t): New typedef.
15214         (operator|, operator|=): Implement for optgroup_flags_t.
15215         (struct dump_file_info): Convert field "alt_flags" to
15216         dump_flags_t.  Convert field "optgroup_flags" to
15217         optgroup_flags_t.
15218         (dump_basic_block): Use dump_flags_t rather than int for param.
15219         (dump_generic_expr_loc): Likewise.
15220         (dump_dec): Likewise.
15221         (dump_register): Convert param "optgroup_flags" to
15222         optgroup_flags_t.
15223         (opt_info_enable_passes): Likewise.
15224         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
15225         than 0.
15226         * gimple-pretty-print.c (debug): Likewise.
15227         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
15228         (merged_store_group::apply_stores): Likewise.
15229         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
15230         * gimple.c (verify_gimple_pp): Likewise.
15231         * graphite-poly.c (print_pbb_body): Likewise.
15232         * passes.c (pass_manager::register_one_dump_file): Convert
15233         local "optgroup_flags" to optgroup_flags_t.
15234         * print-tree.c (print_node): Use TDF_NONE rather than 0.
15235         (debug): Likewise.
15236         (debug_body): Likewise.
15237         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
15238         to optgroup_flags_t.
15239         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
15240         than 0.
15241         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
15242         (convert_mult_to_fma): Likewise.
15243         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
15244         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
15245         * tree-vect-data-refs.c (dump_lower_bound): Convert param
15246         "dump_kind" to dump_flags_t.
15248 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
15250         * config/rs6000/rs6000.c (min, max): Delete.
15252 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
15254         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
15255         -mabi=spe and -mabi=no-spe.
15257 2018-06-08  Martin Liska  <mliska@suse.cz>
15259         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
15260         where we expect an existing summary.
15262 2018-06-08  Martin Liska  <mliska@suse.cz>
15264         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
15265         * ipa-inline.h (estimate_edge_growth): Likewise.
15267 2018-06-08  Martin Liska  <mliska@suse.cz>
15269         * cgraph.c (function_version_hasher::hash): Use
15270         cgraph_node::get_uid ().
15271         (function_version_hasher::equal):
15272         * cgraph.h (cgraph_node::get_uid): New method.
15273         * ipa-inline.c (update_caller_keys): Use
15274         cgraph_node::get_uid ().
15275         (update_callee_keys): Likewise.
15276         * ipa-utils.c (searchc): Likewise.
15277         (ipa_reduced_postorder): Likewise.
15278         * lto-cgraph.c (input_node): Likewise.
15279         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
15280         * symbol-summary.h (symtab_insertion): Likewise.
15281         (symtab_removal): Likewise.
15282         (symtab_duplication): Likewise.
15283         * tree-pretty-print.c (dump_function_header): Likewise.
15284         * tree-sra.c (convert_callers_for_node): Likewise.
15286 2018-06-08  Martin Liska  <mliska@suse.cz>
15288         * cgraph.c (symbol_table::create_edge): Always assign a new
15289         unique number.
15290         (symbol_table::free_edge): Do not recycle numbers.
15291         * cgraph.h (cgraph_edge::get): New method.
15292         * symbol-summary.h (symtab_removal): Use it.
15293         (symtab_duplication): Likewise.
15294         (call_summary::hashable_uid): Remove.
15296 2018-06-08  Martin Liska  <mliska@suse.cz>
15298         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
15299         (initialize_growth_caches): Remove.
15300         (free_growth_caches): Likewise.
15301         (do_estimate_edge_time): Use edge_growth_cache.
15302         (do_estimate_edge_size): Likewise.
15303         (do_estimate_edge_hints): Likewise.
15304         * ipa-inline.c (reset_edge_caches): Likewise.
15305         (recursive_inlining): Likewise.
15306         (inline_small_functions): Likewise.
15307         * ipa-inline.h (initialize_growth_caches): Remove.
15308         (estimate_edge_size): Likewise.
15309         (estimate_edge_time): Likewise.
15310         (estimate_edge_hints): Likewise.
15311         (reset_edge_growth_cache): Likewise.
15312         * symbol-summary.h (call_summary::remove): New method.
15314 2018-06-08  Martin Liska  <mliska@suse.cz>
15316         * ipa-cp.c (class edge_clone_summary): New summary.
15317         (grow_edge_clone_vectors): Remove.
15318         (ipcp_edge_duplication_hook): Remove.
15319         (class edge_clone_summary_t): New call_summary class.
15320         (ipcp_edge_removal_hook): Remove.
15321         (edge_clone_summary_t::duplicate): New function.
15322         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
15323         (create_specialized_node): Likewise.
15324         (ipcp_driver): Initialize edge_clone_summaries and do not
15325         register hooks.
15327 2018-06-08  Martin Liska  <mliska@suse.cz>
15329         * symbol-summary.h (get): New function.
15330         (call_summary::m_initialize_when_cloning): New class member.
15332 2018-06-08  Martin Liska  <mliska@suse.cz>
15334         * cgraph.c (cgraph_node::remove): Do not recycle uid.
15335         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
15336         (symbol_table::allocate_cgraph_symbol): Do not set uid.
15337         * passes.c (uid_hash_t): Record removed_nodes by their uids.
15338         (remove_cgraph_node_from_order): Use the removed_nodes set.
15339         (do_per_function_toporder): Likwise.
15340         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
15341         instead of summary_uid.
15342         (symtab_removal): Likewise.
15343         (symtab_duplication): Likewise.
15345 2018-06-08  Martin Liska  <mliska@suse.cz>
15347         * ipa-cp.c (ipcp_store_bits_results): Use
15348         ipcp_transformation_sum.
15349         (ipcp_store_vr_results): Likewise.
15350         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
15351         to ...
15352         (ipcp_transformation_initialize): ... this.
15353         (ipa_set_node_agg_value_chain):
15354         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
15355         (write_ipcp_transformation_info): Likewise.
15356         (read_ipcp_transformation_info): Likewise.
15357         (ipcp_update_bits): Likewise.
15358         (ipcp_update_vr): Likewise.
15359         (ipcp_transform_function): Likewise.
15360         * ipa-prop.h: Rename ipcp_transformation_summary to
15361         ipcp_transformation.
15362         (class ipcp_transformation_t): New function summary.
15363         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
15364         (ipa_get_agg_replacements_for_node): Likewise.
15366 2018-06-08  Martin Liska  <mliska@suse.cz>
15368         * ipa-pure-const.c (struct funct_state_d): Do it class instead
15369         of struct.
15370         (class funct_state_summary_t): New function_summary class.
15371         (has_function_state): Remove.
15372         (get_function_state): Likewise.
15373         (set_function_state): Likewise.
15374         (add_new_function): Likewise.
15375         (funct_state_summary_t::insert): New function.
15376         (duplicate_node_data): Remove.
15377         (remove_node_data): Remove.
15378         (funct_state_summary_t::duplicate): New function.
15379         (register_hooks): Create new funct_state_summaries.
15380         (pure_const_generate_summary): Use it.
15381         (pure_const_write_summary): Likewise.
15382         (pure_const_read_summary): Likewise.
15383         (propagate_pure_const): Likewise.
15384         (propagate_nothrow): Likewise.
15385         (dump_malloc_lattice): Likewise.
15386         (propagate_malloc): Likewise.
15387         (execute): Do not register hooks, just remove summary
15388         instead.
15389         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
15390         constructor.
15392 2018-06-08  Martin Liska  <mliska@suse.cz>
15394         * ipa-reference.c (remove_node_data): Remove.
15395         (duplicate_node_data): Likewise.
15396         (class ipa_ref_var_info_summary_t): New class.
15397         (class ipa_ref_opt_summary_t): Likewise.
15398         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
15399         (get_reference_optimization_summary): Use
15400         ipa_ref_opt_sum_summaries.
15401         (set_reference_vars_info): Remove.
15402         (set_reference_optimization_summary): Likewise.
15403         (ipa_init): Create summaries.
15404         (init_function_info): Use function summary.
15405         (ipa_ref_opt_summary_t::duplicate): New function.
15406         (ipa_ref_opt_summary_t::remove): New function.
15407         (get_read_write_all_from_node): Fix GNU coding style.
15408         (propagate): Use function summary.
15409         (write_node_summary_p): Fix GNU coding style.
15410         (stream_out_bitmap): Likewise.
15411         (ipa_reference_read_optimization_summary): Use function summary.
15412         (ipa_reference_c_finalize): Do not release hooks.
15414 2018-06-08  Martin Liska  <mliska@suse.cz>
15416         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
15417         (analyze_function_body): Extract multiple calls of get_create.
15418         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
15419         * ipa-inline.c (recursive_inlining): Use ::get method.
15420         * ipa-inline.h (estimate_edge_growth): Likewise.
15422 2018-06-08  Martin Liska  <mliska@suse.cz>
15424         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
15425         HSA_INVALID.
15426         (hsa_function_summary::hsa_function_summary): Use the new enum
15427         value.
15428         (hsa_gpu_implementation_p): Use hsa_summaries::get.
15429         * hsa-gen.c (hsa_get_host_function): Likewise.
15430         (get_brig_function_name): Likewise.
15431         * ipa-hsa.c (process_hsa_functions): Likewise.
15432         (ipa_hsa_write_summary): Likewise.
15433         * symbol-summary.h (symtab_duplication): Use ::get function/
15434         (get): New function.
15436 2018-06-08  Martin Liska  <mliska@suse.cz>
15438         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
15439         of get.
15440         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
15441         (hsa_register_kernel): Likewise.
15442         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
15443         * hsa-gen.c (hsa_get_host_function): Likewise.
15444         (get_brig_function_name): Likewise.
15445         (generate_hsa): Likewise.
15446         (pass_gen_hsail::execute): Likewise.
15447         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
15448         (devirtualization_time_bonus): Likewise.
15449         (ipcp_propagate_stage): Likewise.
15450         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
15451         (edge_set_predicate): Likewise.
15452         (evaluate_conditions_for_known_args): Likewise.
15453         (evaluate_properties_for_edge): Likewise.
15454         (ipa_fn_summary::reset): Likewise.
15455         (ipa_fn_summary_t::duplicate): Likewise.
15456         (dump_ipa_call_summary): Likewise.
15457         (ipa_dump_fn_summary): Likewise.
15458         (analyze_function_body): Likewise.
15459         (compute_fn_summary): Likewise.
15460         (estimate_edge_devirt_benefit): Likewise.
15461         (estimate_edge_size_and_time): Likewise.
15462         (estimate_calls_size_and_time): Likewise.
15463         (estimate_node_size_and_time): Likewise.
15464         (inline_update_callee_summaries): Likewise.
15465         (remap_edge_change_prob): Likewise.
15466         (remap_edge_summaries): Likewise.
15467         (ipa_merge_fn_summary_after_inlining): Likewise.
15468         (ipa_update_overall_fn_summary): Likewise.
15469         (read_ipa_call_summary): Likewise.
15470         (inline_read_section): Likewise.
15471         (write_ipa_call_summary): Likewise.
15472         (ipa_fn_summary_write): Likewise.
15473         (ipa_free_fn_summary): Likewise.
15474         * ipa-hsa.c (process_hsa_functions): Likewise.
15475         (ipa_hsa_write_summary): Likewise.
15476         (ipa_hsa_read_section): Likewise.
15477         * ipa-icf.c (sem_function::merge): Likewise.
15478         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
15479         (do_estimate_edge_time): Likewise.
15480         (estimate_size_after_inlining): Likewise.
15481         (estimate_growth): Likewise.
15482         (growth_likely_positive): Likewise.
15483         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
15484         (inline_call): Likewise.
15485         * ipa-inline.c (caller_growth_limits): Likewise.
15486         (can_inline_edge_p): Likewise.
15487         (can_inline_edge_by_limits_p): Likewise.
15488         (compute_uninlined_call_time): Likewise.
15489         (compute_inlined_call_time): Likewise.
15490         (want_inline_small_function_p): Likewise.
15491         (edge_badness): Likewise.
15492         (update_caller_keys): Likewise.
15493         (update_callee_keys): Likewise.
15494         (recursive_inlining): Likewise.
15495         (inline_small_functions): Likewise.
15496         (inline_to_all_callers_1): Likewise.
15497         (dump_overall_stats): Likewise.
15498         (early_inline_small_functions): Likewise.
15499         (early_inliner): Likewise.
15500         * ipa-inline.h (estimate_edge_growth): Likewise.
15501         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
15502         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15503         * ipa-prop.h (IPA_NODE_REF): Likewise.
15504         (IPA_EDGE_REF): Likewise.
15505         * ipa-pure-const.c (malloc_candidate_p): Likewise.
15506         (propagate_malloc): Likewise.
15507         * ipa-split.c (execute_split_functions): Likewise.
15508         * symbol-summary.h: Rename get to get_create.
15509         (get): Likewise.
15510         (get_create): Likewise.
15511         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
15513 2018-06-08  Martin Liska  <mliska@suse.cz>
15515         * symbol-summary.h (release): Move definition out of class
15516         declaration.
15517         (symtab_removal): Likewise.
15518         (symtab_duplication): Likewise.
15520 2018-06-08  Martin Liska  <mliska@suse.cz>
15522         * symbol-summary.h (function_summary): Move constructor
15523         implementation out of class declaration.
15524         (release): Likewise.
15525         (symtab_insertion): Likewise.
15526         (symtab_removal): Likewise.
15527         (symtab_duplication): Likewise.
15528         (get): Likewise.
15530 2018-06-08  Martin Liska  <mliska@suse.cz>
15532         * Makefile.in: Remove support for MPX (macros, related functions,
15533         fields in cgraph_node, ...).
15534         * builtin-types.def (BT_BND): Likewise.
15535         (BT_FN_BND_CONST_PTR): Likewise.
15536         (BT_FN_CONST_PTR_BND): Likewise.
15537         (BT_FN_VOID_PTR_BND): Likewise.
15538         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
15539         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
15540         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
15541         (expand_builtin_mempcpy_with_bounds): Likewise.
15542         (expand_builtin_memset_with_bounds): Likewise.
15543         (expand_builtin_memset_args): Likewise.
15544         (std_expand_builtin_va_start): Likewise.
15545         (expand_builtin): Likewise.
15546         (expand_builtin_with_bounds): Likewise.
15547         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
15548         (DEF_LIB_BUILTIN_CHKP): Likewise.
15549         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
15550         (DEF_CHKP_BUILTIN): Likewise.
15551         (BUILT_IN_MEMCPY): Likewise.
15552         (BUILT_IN_MEMMOVE): Likewise.
15553         (BUILT_IN_MEMPCPY): Likewise.
15554         (BUILT_IN_MEMSET): Likewise.
15555         (BUILT_IN_STPCPY): Likewise.
15556         (BUILT_IN_STRCAT): Likewise.
15557         (BUILT_IN_STRCHR): Likewise.
15558         (BUILT_IN_STRCPY): Likewise.
15559         (BUILT_IN_STRLEN): Likewise.
15560         (BUILT_IN_MEMCPY_CHK): Likewise.
15561         (BUILT_IN_MEMMOVE_CHK): Likewise.
15562         (BUILT_IN_MEMPCPY_CHK): Likewise.
15563         (BUILT_IN_MEMSET_CHK): Likewise.
15564         (BUILT_IN_STPCPY_CHK): Likewise.
15565         (BUILT_IN_STRCAT_CHK): Likewise.
15566         (BUILT_IN_STRCPY_CHK): Likewise.
15567         * calls.c (store_bounds): Likewise.
15568         (emit_call_1): Likewise.
15569         (special_function_p): Likewise.
15570         (maybe_warn_nonstring_arg): Likewise.
15571         (initialize_argument_information): Likewise.
15572         (finalize_must_preallocate): Likewise.
15573         (compute_argument_addresses): Likewise.
15574         (expand_call): Likewise.
15575         * cfgexpand.c (expand_call_stmt): Likewise.
15576         (expand_return): Likewise.
15577         (expand_gimple_stmt_1): Likewise.
15578         (pass_expand::execute): Likewise.
15579         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
15580         (cgraph_node::remove): Likewise.
15581         (cgraph_node::dump): Likewise.
15582         (cgraph_node::verify_node): Likewise.
15583         * cgraph.h (chkp_function_instrumented_p): Likewise.
15584         (symtab_node::get_alias_target): Likewise.
15585         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
15586         (cgraph_local_p): Likewise.
15587         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
15588         (cgraph_edge::rebuild_references): Likewise.
15589         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
15590         (walk_polymorphic_call_targets): Likewise.
15591         (cgraph_node::expand_thunk): Likewise.
15592         (symbol_table::output_weakrefs): Likewise.
15593         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
15594         (ix86_handle_option): Likewise.
15595         * config/i386/constraints.md: Likewise.
15596         * config/i386/i386-builtin-types.def (BND): Likewise.
15597         (VOID): Likewise.
15598         (PVOID): Likewise.
15599         (ULONG): Likewise.
15600         * config/i386/i386-builtin.def (BDESC_END): Likewise.
15601         (BDESC_FIRST): Likewise.
15602         (BDESC): Likewise.
15603         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
15604         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
15605         * config/i386/i386.c (enum reg_class): Likewise.
15606         (ix86_target_string): Likewise.
15607         (ix86_option_override_internal): Likewise.
15608         (ix86_conditional_register_usage): Likewise.
15609         (ix86_valid_target_attribute_inner_p): Likewise.
15610         (ix86_set_indirect_branch_type): Likewise.
15611         (ix86_set_current_function): Likewise.
15612         (ix86_function_arg_regno_p): Likewise.
15613         (init_cumulative_args): Likewise.
15614         (ix86_function_arg_advance): Likewise.
15615         (ix86_function_arg): Likewise.
15616         (ix86_pass_by_reference): Likewise.
15617         (ix86_function_value_regno_p): Likewise.
15618         (ix86_function_value_1): Likewise.
15619         (ix86_function_value_bounds): Likewise.
15620         (ix86_return_in_memory): Likewise.
15621         (ix86_setup_incoming_vararg_bounds): Likewise.
15622         (ix86_va_start): Likewise.
15623         (indirect_thunk_need_prefix): Likewise.
15624         (print_reg): Likewise.
15625         (ix86_print_operand): Likewise.
15626         (ix86_expand_call): Likewise.
15627         (ix86_output_function_return): Likewise.
15628         (reg_encoded_number): Likewise.
15629         (BDESC_VERIFYS): Likewise.
15630         (ix86_init_mpx_builtins): Likewise.
15631         (ix86_init_builtins): Likewise.
15632         (ix86_emit_cmove): Likewise.
15633         (ix86_emit_move_max): Likewise.
15634         (ix86_expand_builtin): Likewise.
15635         (ix86_builtin_mpx_function): Likewise.
15636         (ix86_get_arg_address_for_bt): Likewise.
15637         (ix86_load_bounds): Likewise.
15638         (ix86_store_bounds): Likewise.
15639         (ix86_load_returned_bounds): Likewise.
15640         (ix86_store_returned_bounds): Likewise.
15641         (ix86_class_likely_spilled_p): Likewise.
15642         (ix86_hard_regno_mode_ok): Likewise.
15643         (x86_order_regs_for_local_alloc): Likewise.
15644         (ix86_mitigate_rop): Likewise.
15645         (ix86_bnd_prefixed_insn_p): Likewise.
15646         (ix86_mpx_bound_mode): Likewise.
15647         (ix86_make_bounds_constant): Likewise.
15648         (ix86_initialize_bounds): Likewise.
15649         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
15650         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
15651         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
15652         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
15653         (TARGET_CHKP_BOUND_MODE): Likewise.
15654         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
15655         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
15656         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
15657         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
15658         * config/i386/i386.h (TARGET_MPX): Likewise.
15659         (TARGET_MPX_P): Likewise.
15660         (VALID_BND_REG_MODE): Likewise.
15661         (FIRST_BND_REG): Likewise.
15662         (LAST_BND_REG): Likewise.
15663         (enum reg_class): Likewise.
15664         (BND_REG_P): Likewise.
15665         (BND_REGNO_P): Likewise.
15666         (BNDmode): Likewise.
15667         (ADJUST_INSN_LENGTH): Likewise.
15668         * config/i386/i386.md: Likewise.
15669         * config/i386/i386.opt: Likewise.
15670         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
15671         (defined): Likewise.
15672         (LINK_MPX): Likewise.
15673         (MPX_SPEC): Likewise.
15674         (LIBMPX_SPEC): Likewise.
15675         (LIBMPXWRAPPERS_SPEC): Likewise.
15676         (CHKP_SPEC): Likewise.
15677         * config/i386/predicates.md: Likewise.
15678         * dbxout.c (dbxout_type): Likewise.
15679         * doc/extend.texi: Likewise.
15680         * doc/invoke.texi: Likewise.
15681         * doc/md.texi: Likewise.
15682         * doc/tm.texi: Likewise.
15683         * doc/tm.texi.in: Likewise.
15684         * dwarf2out.c (is_base_type): Likewise.
15685         (gen_formal_types_die): Likewise.
15686         (gen_subprogram_die): Likewise.
15687         (gen_type_die_with_usage): Likewise.
15688         (gen_decl_die): Likewise.
15689         (dwarf2out_late_global_decl): Likewise.
15690         * expr.c (expand_assignment): Likewise.
15691         (emit_storent_insn): Likewise.
15692         (store_expr_with_bounds): Likewise.
15693         (store_expr): Likewise.
15694         (expand_expr_real_1): Likewise.
15695         * expr.h (store_expr_with_bounds): Likewise.
15696         * function.c (use_register_for_decl): Likewise.
15697         (struct bounds_parm_data): Likewise.
15698         (assign_parms_augmented_arg_list): Likewise.
15699         (assign_parm_find_entry_rtl): Likewise.
15700         (assign_parm_is_stack_parm): Likewise.
15701         (assign_parm_load_bounds): Likewise.
15702         (assign_bounds): Likewise.
15703         (assign_parms): Likewise.
15704         (expand_function_start): Likewise.
15705         * gcc.c (CHKP_SPEC): Likewise.
15706         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
15707         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
15708         (wrestrict_dom_walker::check_call): Likewise.
15709         * gimple.c (gimple_build_call_from_tree): Likewise.
15710         * gimple.h (enum gf_mask): Likewise.
15711         (gimple_call_with_bounds_p): Likewise.
15712         (gimple_call_set_with_bounds): Likewise.
15713         * gimplify.c (gimplify_init_constructor): Likewise.
15714         * ipa-cp.c (initialize_node_lattices): Likewise.
15715         (propagate_constants_across_call): Likewise.
15716         (find_more_scalar_values_for_callers_subset): Likewise.
15717         * ipa-hsa.c (process_hsa_functions): Likewise.
15718         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
15719         * ipa-icf.c (sem_function::merge): Likewise.
15720         * ipa-inline.c (early_inliner): Likewise.
15721         * ipa-pure-const.c (warn_function_noreturn): Likewise.
15722         (warn_function_cold): Likewise.
15723         (propagate_pure_const): Likewise.
15724         * ipa-ref.h (enum GTY): Likewise.
15725         * ipa-split.c (find_retbnd): Likewise.
15726         (consider_split): Likewise.
15727         (split_function): Likewise.
15728         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
15729         * ipa.c (walk_polymorphic_call_targets): Likewise.
15730         (symbol_table::remove_unreachable_nodes): Likewise.
15731         (process_references): Likewise.
15732         (cgraph_build_static_cdtor_1): Likewise.
15733         * lto-cgraph.c (lto_output_node): Likewise.
15734         (output_refs): Likewise.
15735         (compute_ltrans_boundary): Likewise.
15736         (input_overwrite_node): Likewise.
15737         (input_node): Likewise.
15738         (input_cgraph_1): Likewise.
15739         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
15740         * passes.c (pass_manager::execute_early_local_passes): Likewise.
15741         (class pass_chkp_instrumentation_passes): Likewise.
15742         (make_pass_chkp_instrumentation_passes): Likewise.
15743         * passes.def: Likewise.
15744         * rtl.h (struct GTY): Likewise.
15745         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
15746         * stor-layout.c (layout_type): Likewise.
15747         * symtab.c: Likewise.
15748         * target.def: Likewise.
15749         * targhooks.c (default_chkp_bound_type): Likewise.
15750         (default_chkp_bound_mode): Likewise.
15751         (default_builtin_chkp_function): Likewise.
15752         (default_chkp_function_value_bounds): Likewise.
15753         (default_chkp_make_bounds_constant): Likewise.
15754         (default_chkp_initialize_bounds): Likewise.
15755         * targhooks.h (default_chkp_bound_type): Likewise.
15756         (default_chkp_bound_mode): Likewise.
15757         (default_builtin_chkp_function): Likewise.
15758         (default_chkp_function_value_bounds): Likewise.
15759         (default_chkp_make_bounds_constant): Likewise.
15760         (default_chkp_initialize_bounds): Likewise.
15761         * toplev.c (compile_file): Likewise.
15762         (process_options): Likewise.
15763         * tree-core.h (DEF_BUILTIN): Likewise.
15764         (DEF_BUILTIN_CHKP): Likewise.
15765         * tree-inline.c (declare_return_variable): Likewise.
15766         (remap_gimple_stmt): Likewise.
15767         (copy_bb): Likewise.
15768         (initialize_inlined_parameters): Likewise.
15769         (expand_call_inline): Likewise.
15770         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
15771         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
15772         (make_pass_ipa_chkp_produce_thunks): Likewise.
15773         (make_pass_chkp): Likewise.
15774         (make_pass_chkp_opt): Likewise.
15775         (make_pass_chkp_instrumentation_passes): Likewise.
15776         * tree-pretty-print.c (dump_generic_node): Likewise.
15777         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
15778         * tree-ssa-dce.c (propagate_necessity): Likewise.
15779         (eliminate_unnecessary_stmts): Likewise.
15780         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
15781         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
15782         * tree-ssa-sccvn.h: Likewise.
15783         * tree-ssa-strlen.c (get_string_length): Likewise.
15784         (valid_builtin_call): Likewise.
15785         (adjust_last_stmt): Likewise.
15786         (handle_builtin_strchr): Likewise.
15787         (handle_builtin_strcpy): Likewise.
15788         (handle_builtin_stxncpy): Likewise.
15789         (handle_builtin_memcpy): Likewise.
15790         (handle_builtin_strcat): Likewise.
15791         (strlen_check_and_optimize_stmt): Likewise.
15792         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
15793         * tree-streamer-in.c: Likewise.
15794         * tree-streamer.c (record_common_node): Likewise.
15795         * tree.c (tree_code_size): Likewise.
15796         (wide_int_to_tree_1): Likewise.
15797         (type_contains_placeholder_1): Likewise.
15798         (build_common_tree_nodes): Likewise.
15799         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
15800         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
15801         (POINTER_BOUNDS_P): Likewise.
15802         (BOUNDED_TYPE_P): Likewise.
15803         (BOUNDED_P): Likewise.
15804         (CALL_WITH_BOUNDS_P): Likewise.
15805         (pointer_bounds_type_node): Likewise.
15806         * value-prof.c (gimple_ic): Likewise.
15807         * var-tracking.c (vt_add_function_parameters): Likewise.
15808         * varasm.c (make_decl_rtl): Likewise.
15809         (assemble_start_function): Likewise.
15810         (output_constant): Likewise.
15811         (maybe_assemble_visibility): Likewise.
15812         * varpool.c (ctor_for_folding): Likewise.
15813         * chkp-builtins.def: Remove.
15814         * ipa-chkp.c: Remove.
15815         * ipa-chkp.h: Remove.
15816         * rtl-chkp.c: Remove.
15817         * rtl-chkp.h: Remove.
15818         * tree-chkp-opt.c: Remove.
15819         * tree-chkp.c: Remove.
15820         * tree-chkp.h: Remove.
15822 2018-06-07  Carl Love  <cel@us.ibm.com>
15824         * config/rs6000/vsx.md (vextract_fp_from_shorth,
15825         vextract_fp_from_shortl): Add BE support.
15827 2018-06-07  Paul Koning  <ni1d@arrl.net>
15829         * compare-elim.c (try_merge_compare): Don't merge compare if
15830         address contains a side effect.
15831         (try_eliminate_compare): Likewise.
15833 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
15835         * config.gcc: Support "tremont".
15836         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
15837         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15838         PROCESSOR_TREMONT.
15839         * config/i386/i386.c (m_TREMONT): Define.
15840         (processor_target_table): Add "tremont".
15841         (PTA_TREMONT): Define.
15842         (ix86_lea_outperforms): Add TARGET_TREMONT.
15843         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
15844         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
15845         and M_INTEL_GOLDMONT_PLUS.
15846         (fold_builtin_cpu): Add "tremont".
15847         (ix86_add_stmt_cost): Add TARGET_TREMONT.
15848         (ix86_option_override_internal): Add "tremont".
15849         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
15850         (processor_type): Add PROCESSOR_TREMONT.
15851         * config/i386/x86-tune.def: Add m_TREMONT.
15852         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
15854 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15856         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
15857         symbol defined for msp430i* devices to be lower case.
15859 2018-06-07  Richard Biener  <rguenther@suse.de>
15861         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
15862         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
15863         Properly wrap signed arithmetic if overflow wraps.
15865 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
15867         PR tree-optimization/69615
15868         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
15869         of a cast from a same precision integral SSA_NAME in a bb dominated
15870         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
15871         cast to utype if rhs2 has already a compatible type.
15873 2018-06-07  Richard Biener  <rguenther@suse.de>
15875         PR tree-optimization/85935
15876         * graphite-scop-detection.c (find_params_in_bb): Analyze
15877         condition operands with respect to the correct loop.  Assert
15878         the analysis doesn't fail.
15880 2018-06-04  Carl Love  <cel@us.ibm.com>
15882         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
15883         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
15884         as it is slightly cheaper.
15885         (first_match_or_eos_index_<mode>):
15886         Calculate index using natural element order.
15887         (first_match_index_<mode>):
15888         Calculate index using natural element order.
15889         (first_match_or_eos_index_<mode>):
15890         Calculate index using natural order.
15891         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
15892         for BE and LE modes.
15893         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
15894         P9V_BUILTIN_VCLZLSBB_V16QI.
15895         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
15896         specific.
15898 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15900         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
15901         indentation and line wrap for many prototypes.  Add missing
15902         @smallexample directives around block of prototypes for vec_xl and
15903         vec_xst.
15905 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
15907         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
15908         track if we pass or return IEEE 128-bit floating point.
15909         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
15910         C++ mangling that is compatible with GCC 8.1.
15911         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
15912         (init_cumulative_args): Note if we pass or return IEEE 128-bit
15913         floating point types.
15914         (rs6000_function_arg_advance_1): Likewise.
15915         (rs6000_mangle_type): Optionally generate mangled names that match
15916         what GCC 8.1 generated for IEEE 128-bit floating point types.
15917         (rs6000_globalize_decl_name): If we have an external function that
15918         passes or returns IEEE 128-bit types, generate a weak reference
15919         from the mangled name used in GCC 8.1 to the current mangled
15920         name.
15921         (rs6000_init_builtins): Make __ibm128 use the long double type if
15922         long double is IBM extended double.  Make __float128 use the long
15923         double type if long double is IEEE 128-bit.
15925         PR target/85657
15926         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
15927         macro for __ibm128 built-in functions.
15928         (PACK_IF): Add __ibm128 pack/unpack functions.
15929         (UNPACK_IF): Likewise.
15930         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
15931         enable long double built-in functions if long double is IEEE
15932         128-bit floating point.
15933         (rs6000_invalid_builtin): Update long double built-in function
15934         error message.
15935         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
15936         functions, adjust the built-in function to use the long double
15937         built-in function if __ibm128 and long double are the same type.
15938         * doc/extend.texi (PowerPC builtins): Update documention for
15939         __builtin_{,un}pack_longdouble.  Add documentation for
15940         __builtin_{,un}pack_ibm128.
15942 2018-06-06  Jim Wilson  <jimw@sifive.com>
15944         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
15945         (struct machine_function): New field interrupt_mode.
15946         (riscv_handle_type_attribute): New function.  Add forward declaration.
15947         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
15948         (riscv_expand_epilogue): Check interrupt_mode field.
15949         (riscv_set_current_function): Check interrupt attribute args and
15950         set interrupt_mode field.
15951         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
15952         (riscv_sret, riscv_uret): New.
15953         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
15954         new arguments to interrupt attribute.
15956 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
15958         PR target/63177
15959         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
15960         Don't handle -mcpu=power8 if -mpower9-vector is also used.
15962 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15964         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
15965         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
15966         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
15967         several redundant entries.
15969 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
15971         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
15972         type from "rtx" to "rtx_insn *".
15973         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
15974         for local "call_insn", removing cast.
15975         (ix86_expand_call): Likewise, introducing a "call_insn" local.
15977 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
15979         PR tree-optimization/86066
15980         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
15981         for BIT_INSERT_EXPR stores.
15983 2018-06-06  Richard Biener  <rguenther@suse.de>
15985         PR tree-optimization/86062
15986         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
15987         component refs ontop
15988         of to be offsetted base.
15990 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15992         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
15993         to be static and remove check on interrupt attribute name.
15995 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15997         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
15998         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
16000 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
16002         PR target/79924
16003         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
16004         second argument.
16005         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
16006         Remove second argument, change how error is called.
16007         (aarch64_layout_arg): Remove second argument from
16008         aarch64_err_no_fpadvsimd call.
16009         (aarch64_init_cumulative_args): Ditto.
16010         (aarch64_gimplify_va_arg_expr): Ditto.
16011         * config/aarch64/aarch64.md (mov<mode>): Ditto.
16013 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
16015         * config/i386/i386.md (simple_return_indirect_internal): New expander.
16016         (*simple_return_indirect_internal<mode>): Rename from
16017         simple_return_indirect_internal.  Use W mode iterator.
16018         (rstorssp): New expander.
16019         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
16020         (clrssbsy): New expander.
16021         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
16023 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16025         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
16026         __typeof__.
16027         (cmse_check_pointed_object): Likewise.
16029 2018-06-05  Martin Liska  <mliska@suse.cz>
16031         PR gcov-profile/47618
16032         * doc/invoke.texi: Document how -fprofile-dir format
16033         is extended.
16035 2018-06-05  Richard Biener  <rguenther@suse.de>
16037         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
16038         removal pretend DOM info isn't available so we do not update
16039         it and only remove edges, not dominated blocks.  Actually free
16040         DOM info in case we removed something.  Remove unreachable blocks.
16041         (mfb_keep_latches): Work with either DOM info or marked backedges.
16042         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
16043         first.  Mark backedges if DOM info isn't available.
16044         (Re-)compute DOM info after cleanup_control_flow_pre.
16046 2018-06-05  Richard Biener  <rguenther@suse.de>
16048         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
16049         (locus_discrim_hasher::hash): Adjust.
16050         (locus_discrim_hasher::equal): Likewise.
16051         (next_discriminator_for_locus): Work on line directly.
16052         (same_line_p): Pass in expanded locus1 as well.
16053         (assign_discriminators): Avoid redundant location expansions.
16055 2018-06-05  Richard Biener  <rguenther@suse.de>
16057         PR tree-optimization/86046
16058         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
16059         if required after clearing TREE_ADDRESSABLE.
16061 2018-06-05  Richard Biener  <rguenther@suse.de>
16063         PR tree-optimization/86047
16064         * tree-ssa-loop.c (for_each_index): Glob handling of all
16065         decls and constants and really handle all of them.
16067 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16069         PR target/81497
16070         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
16071         qualifier_void_pointer and qualifier_const_void_pointer.
16072         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
16073         (arm_init_builtins): Handle the above.
16074         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
16075         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
16076         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
16077         void intrinsics.
16079 2018-06-05  Martin Liska  <mliska@suse.cz>
16081         * auto-profile.c (read_autofdo_file): Do not use
16082         gcov_ctr_summary struct.
16083         (afdo_callsite_hot_enough_for_early_inline): Likewise.
16084         * coverage.c (struct counts_entry): Likewise.
16085         (read_counts_file): Read just single summary entry.
16086         (get_coverage_counts): Use gcov_summary struct.
16087         * coverage.h (get_coverage_counts): Likewise.
16088         * gcov-dump.c (dump_working_sets): Likewise.
16089         (tag_summary): Dump just single summary.
16090         * gcov-io.c (gcov_write_summary): Write just histogram
16091         summary.
16092         (gcov_read_summary): Read just single summary.
16093         (compute_working_sets): Use gcov_summary struct.
16094         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
16095         of GCOV_COUNTERS_SUMMABLE.
16096         (GCOV_COUNTERS_SUMMABLE): Remove.
16097         (GCOV_FIRST_VALUE_COUNTER): Replace with
16098         GCOV_COUNTER_V_INTERVAL.
16099         (struct gcov_ctr_summary): Remove.
16100         (struct gcov_summary): Directly use fields of former
16101         gcov_ctr_summary.
16102         (compute_working_sets): Use gcov_summary struct.
16103         * gcov.c (read_count_file): Do not use ctrs fields.
16104         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
16105         struct.
16106         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
16107         struct.
16108         * profile.c: Likewise.
16109         * profile.h: Likewise.
16111 2018-06-05  Martin Liska  <mliska@suse.cz>
16113         PR gcov-profile/84846
16114         * gcov.c (output_lines): Print working directory only
16115         in intermediate format.
16117 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
16119         * config/s390/s390-builtin-types.def: Add void function type.
16120         * config/s390/s390-builtins.def: Use the function type for the
16121         tbeginc builtin.
16123 2018-06-04  Jim Wilson  <jimw@sifive.com>
16125         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
16126         to int.
16127         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
16128         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
16129         handle EH_RETURN_DATA_REGNO registers properly.
16130         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
16131         (riscv_expand_epilogue): Update comment.  Change argument name and
16132         type.  Update code to use new name and type.  Pass new args to
16133         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
16134         EXCEPTION_RETURN.
16135         * config/riscv/riscv.md (NORMAL_RETURN): New.
16136         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
16137         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
16138         (eh_return): Call gen_eh_return_internal and emit barrier.
16139         (eh_return_internal): Call riscv_expand_epilogue.
16141 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
16143         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
16144         bit_insertion field and declare can_be_merged_into method.
16145         (merged_store_group::can_be_merged_into): New method.
16146         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
16147         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
16148         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
16150 2018-06-04  Richard Biener  <rguenther@suse.de>
16152         PR tree-optimization/85955
16153         * builtins.c (fold_builtin_sincos): Convert pointers to
16154         destination to appropriate type before dereferencing.
16156 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
16158         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
16160 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
16162         * expr.c (expand_expr_real_1): Force the operand into memory if
16163         its TYPE_MODE is BLKmode and if there is no integer mode for
16164         the number of bits being extracted.
16166 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
16168         PR target/85832
16169         PR target/86036
16170         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
16171         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
16173 2018-06-04  Richard Biener  <rguenther@suse.de>
16175         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
16176         (cleanup_tree_cfg_noloop): ... single caller.  Do
16177         start_recording_case_labels later.
16179 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
16181         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
16182         to _IMMINTRIN_H_INCLUDED.
16183         * config/i386/pconfigintrin.h: Ditto.
16184         * config/i386/waitpkgintrin.h: Ditto.
16185         * config/i386/immintrin.h: Add includes for sgxintrin.h,
16186         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
16187         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
16188         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
16189         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
16190         waitpkgintrin.h and cldemoteintrin.h.
16192 2018-06-04  Richard Biener  <rguenther@suse.de>
16194         PR tree-optimization/86038
16195         * tracer.c (find_best_successor): Check probability for
16196         being initialized, bail out if not.
16198 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
16200         PR target/86003
16201         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
16202         of bits to ignore when comparing architectures.
16204 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
16206         PR tree-optimization/69615
16207         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
16208         maximum or minimum of the type, try to merge it also as if
16209         range1 is + [-, x - 1] or + [x + 1, -].
16211         PR c++/86025
16212         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
16214 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
16216         PR tree-optimization/86034
16217         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
16218         the unsigned bitfield type in a bit insertion sequence if it does not
16219         have a larger precision than the bitfield size.
16220         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
16222 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
16224         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
16226 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
16228         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
16229         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
16230         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
16231         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
16233 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
16235         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
16236         Disable -fdelete-null-pointer-checks for ELF toolchain.
16238 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
16239             Kito Cheng  <kito.cheng@gmail.com>
16241         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
16242         (nds32le-*-*, nds32be-*-*): Integrate checking process.
16243         (nds32*-*-*): Add glibc and uclibc conditions.
16244         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
16245         (TARGET_EXCEPT_UNWIND_INFO): Define.
16246         * config/nds32/elf.h: New file.
16247         * config/nds32/linux.h: New file.
16248         * config/nds32/nds32-elf.opt: New file.
16249         * config/nds32/nds32-linux.opt: New file.
16250         * config/nds32/nds32-fp-as-gp.c
16251         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
16252         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
16253         TARGET_LINUX_ABI.
16254         (nds32_asm_file_end): Ditto.
16255         (nds32_print_operand): Ditto.
16256         (nds32_insert_attributes): Ditto.
16257         (nds32_init_libfuncs): New function.
16258         (TARGET_HAVE_TLS): Define.
16259         (TARGET_INIT_LIBFUNCS): Define.
16260         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
16261         spec content.
16262         (TARGET_ELF): Apply different mcmodel setting.
16263         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
16264         been migrated into elf.h and linux.h files.
16265         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
16266         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
16267         (mcmodel): The content has been migrated into nds32-elf.opt and
16268         nds32-linux.opt files.
16269         * config/nds32/t-elf: New file.
16270         * config/nds32/t-linux: New file.
16272 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
16273             Shiva Chen  <shiva0217@gmail.com>
16275         * config/nds32/constants.md (unspec_volatile_element): Add
16276         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
16277         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
16278         optimization.
16279         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
16280         (make_pass_nds32_fp_as_gp): Declare.
16281         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
16282         optmization pass.
16283         (nds32_asm_function_end_prologue): Remove unused asm output.
16284         (nds32_asm_function_begin_epilogue): Remove unused asm output.
16285         (nds32_asm_file_start): Output necessary fp_as_gp information.
16286         (nds32_option_override): Adjust register usage.
16287         (nds32_expand_prologue): Consider fp_as_gp situation.
16288         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
16289         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
16290         (epilogue): Ditto.
16291         (return): Ditto.
16292         (simple_return): Ditto.
16293         (omit_fp_begin): Output special directive for fp_as_gp.
16294         (omit_fp_end): Output special directive for fp_as_gp.
16295         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
16296         mforbid-fp-as-gp): New options.
16298 2018-06-01  Mark Wielaard  <mark@klomp.org>
16300         * dwarf2out.c (dwarf2out_finish): Remove generation of
16301         DW_AT_loclists_base.
16303 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
16305         * gimple-ssa-store-merging.c: Include gimple-fold.h.
16306         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
16307         (struct merged_store_group): Add bit_insertion field.
16308         (dump_char_array): Use standard hexadecimal format.
16309         (merged_store_group::merged_store_group): Set bit_insertion to false.
16310         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
16311         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
16312         also print the mask in the dump file.
16313         (pass_store_merging::gate): Minor tweak.
16314         (imm_store_chain_info::coalesce_immediate): Fix wrong association
16315         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
16316         stores with INTEGER_CST stores.
16317         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
16318         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
16319         and use it throughout.  Generate bit insertion sequences if need be.
16320         (pass_store_merging::process_store): Remove redundant condition.
16321         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
16323 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
16325         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
16326         the 128-bit floating point types.  Fix function comment.
16328 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16330         * config/aarch64/aarch64-simd.md
16331         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
16332         mnemonics.
16333         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
16334         mnemonics.
16336 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
16338         PR tree-optimization/85989
16339         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
16340         variable.
16341         (backprop::intersect_uses): Check it when deciding whether this
16342         is a backedge reference.
16343         (backprop::process_block): Add each phi to m_visited_phis
16344         after visiting it, then clear it at the end.
16346 2018-06-01  Richard Biener  <rguenther@suse.de>
16348         * tree-vectorizer.h (vect_dr_stmt): New function.
16349         (vect_get_load_cost): Adjust.
16350         (vect_get_store_cost): Likewise.
16351         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
16352         Use vect_dr_stmt instead of DR_SMTT.
16353         (vect_record_base_alignments): Likewise.
16354         (vect_calculate_target_alignment): Likewise.
16355         (vect_compute_data_ref_alignment): Likewise and make static.
16356         (vect_update_misalignment_for_peel): Likewise.
16357         (vect_verify_datarefs_alignment): Likewise.
16358         (vector_alignment_reachable_p): Likewise.
16359         (vect_get_data_access_cost): Likewise.  Pass down
16360         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
16361         (vect_get_peeling_costs_all_drs): Likewise.
16362         (vect_peeling_hash_get_lowest_cost): Likewise.
16363         (vect_enhance_data_refs_alignment): Likewise.
16364         (vect_find_same_alignment_drs): Likewise.
16365         (vect_analyze_data_refs_alignment): Likewise.
16366         (vect_analyze_group_access_1): Likewise.
16367         (vect_analyze_group_access): Likewise.
16368         (vect_analyze_data_ref_access): Likewise.
16369         (vect_analyze_data_ref_accesses): Likewise.
16370         (vect_vfa_segment_size): Likewise.
16371         (vect_small_gap_p): Likewise.
16372         (vectorizable_with_step_bound_p): Likewise.
16373         (vect_prune_runtime_alias_test_list): Likewise.
16374         (vect_analyze_data_refs): Likewise.
16375         (vect_supportable_dr_alignment): Likewise.
16376         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
16377         (vect_gen_prolog_loop_niters): Likewise.
16378         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
16379         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
16380         modify DR_STMT.
16381         (vect_recog_mask_conversion_pattern): Likewise.
16382         (vect_try_gather_scatter_pattern): Likewise.
16383         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
16384         to vect_get_store_cost.
16385         (vect_get_store_cost): Get stmt_info instead of DR.
16386         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
16387         (vect_get_load_cost): Get stmt_info instead of DR.
16389 2018-06-01  Richard Biener  <rguenther@suse.de>
16391         PR middle-end/86017
16392         * gimple-fold.c (var_decl_component_p): Also allow offsetted
16393         vars wrapped in MEM_REFs.
16395 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
16397         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
16398         Fix subreg tests so that we only return a choice between
16399         GENERAL_REGS and FP_REGS if the original classes included both.
16401 2018-06-01  Richard Biener  <rguenther@suse.de>
16403         PR ipa/85960
16404         * tree-ssa-structalias.c (get_function_part_constraint):
16405         Handle NULL fi->decl.
16406         (find_func_aliases_for_call): Properly handle indirect
16407         fi from direct call.
16408         (find_func_clobbers): Likewise.
16409         (ipa_pta_execute): Likewise.
16410         (create_variable_info_for): For functions that are ifunc_resolver
16411         resolve to a varinfo that contains the result of the resolver call.
16412         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
16413         aliases.
16415 2018-05-31  Michael Collison  <michael.collison@arm.com>
16417         * config/aarch64/aarch64.md:
16418         (*fix_to_zero_extenddfdi2): New pattern.
16419         * gcc.target/aarch64/fix_extend1.c: New testcase.
16421 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
16423         PR middle-end/78809
16424         PR middle-end/83026
16425         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
16426         and BUILT_IN_STRNCMP_EQ.
16427         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
16428         BUILT_IN_STRNCMP_EQ.
16429         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
16430         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
16431         (gimple_fold_builtin): Likewise.
16432         * tree-ssa-strlen.c (compute_string_length): New function.
16433         (determine_min_obsize): New function.
16434         (handle_builtin_string_cmp): New function to handle calls to
16435         string compare functions.
16436         (strlen_optimize_stmt): Add handling to builtin string compare
16437         calls.
16438         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
16439         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
16440         * tree.c (build_common_builtin_nodes): Add new defines of
16441         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
16443 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
16445         PR target/85984
16446         * bb-reorder.c (pass_partition_blocks::gate): Return false for
16447         functions with naked attribute.
16449 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
16451         * config/i386/sse.md (avx_vec_concat<mode>):
16452         Substitute concat_tg_mode mode attribute with xtg_mode.
16453         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
16454         (concat_tg_mode): Remove mode attribute.
16456 2018-05-31  Martin Sebor  <msebor@redhat.com>
16458         PR c/82063
16459         * calls.c (alloc_max_size): Correct a logic error/typo.
16460         Treat excessive arguments as infinite.  Warn for invalid arguments.
16461         * doc/invoke.texi (-Walloc-size-larger-than): Update.
16463 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
16465         PR target/85829
16466         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
16467         and movx for Haswell.
16469 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
16470             Cesar Philippidis  <cesar@codesourcery.com>
16472         PR middle-end/85879
16473         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
16474         when emitting error on private/firstprivate reductions.
16475         * omp-low.c (lower_omp_target): Avoid reference-type processing
16476         on pointers for firstprivate clause.
16478 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
16480         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
16481         (st1x2): Likewise.
16482         (st1x3): Likewise.
16483         * config/aarch64/aarch64-simd.md
16484         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
16485         (aarch64_ld1_x3_<mode>): Likewise
16486         (aarch64_st1x2<VALLDIF:mode>): Likewise
16487         (aarch64_st1_x2_<mode>): Likewise
16488         (aarch64_st1x3<VALLDIF:mode>): Likewise
16489         (aarch64_st1_x3_<mode>): Likewise
16490         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
16491         (vld1_s8_x3): Likewise.
16492         (vld1_u16_x3): Likewise.
16493         (vld1_s16_x3): Likewise.
16494         (vld1_u32_x3): Likewise.
16495         (vld1_s32_x3): Likewise.
16496         (vld1_u64_x3): Likewise.
16497         (vld1_s64_x3): Likewise.
16498         (vld1_f16_x3): Likewise.
16499         (vld1_f32_x3): Likewise.
16500         (vld1_f64_x3): Likewise.
16501         (vld1_p8_x3): Likewise.
16502         (vld1_p16_x3): Likewise.
16503         (vld1_p64_x3): Likewise.
16504         (vld1q_u8_x3): Likewise.
16505         (vld1q_s8_x3): Likewise.
16506         (vld1q_u16_x3): Likewise.
16507         (vld1q_s16_x3): Likewise.
16508         (vld1q_u32_x3): Likewise.
16509         (vld1q_s32_x3): Likewise.
16510         (vld1q_u64_x3): Likewise.
16511         (vld1q_s64_x3): Likewise.
16512         (vld1q_f16_x3): Likewise.
16513         (vld1q_f32_x3): Likewise.
16514         (vld1q_f64_x3): Likewise.
16515         (vld1q_p8_x3): Likewise.
16516         (vld1q_p16_x3): Likewise.
16517         (vld1q_p64_x3): Likewise.
16518         (vst1_s64_x2): Likewise.
16519         (vst1_u64_x2): Likewise.
16520         (vst1_f64_x2): Likewise.
16521         (vst1_s8_x2): Likewise.
16522         (vst1_p8_x2): Likewise.
16523         (vst1_s16_x2): Likewise.
16524         (vst1_p16_x2): Likewise.
16525         (vst1_s32_x2): Likewise.
16526         (vst1_u8_x2): Likewise.
16527         (vst1_u16_x2): Likewise.
16528         (vst1_u32_x2): Likewise.
16529         (vst1_f16_x2): Likewise.
16530         (vst1_f32_x2): Likewise.
16531         (vst1_p64_x2): Likewise.
16532         (vst1q_s8_x2): Likewise.
16533         (vst1q_p8_x2): Likewise.
16534         (vst1q_s16_x2): Likewise.
16535         (vst1q_p16_x2): Likewise.
16536         (vst1q_s32_x2): Likewise.
16537         (vst1q_s64_x2): Likewise.
16538         (vst1q_u8_x2): Likewise.
16539         (vst1q_u16_x2): Likewise.
16540         (vst1q_u32_x2): Likewise.
16541         (vst1q_u64_x2): Likewise.
16542         (vst1q_f16_x2): Likewise.
16543         (vst1q_f32_x2): Likewise.
16544         (vst1q_f64_x2): Likewise.
16545         (vst1q_p64_x2): Likewise.
16546         (vst1_s64_x3): Likewise.
16547         (vst1_u64_x3): Likewise.
16548         (vst1_f64_x3): Likewise.
16549         (vst1_s8_x3): Likewise.
16550         (vst1_p8_x3): Likewise.
16551         (vst1_s16_x3): Likewise.
16552         (vst1_p16_x3): Likewise.
16553         (vst1_s32_x3): Likewise.
16554         (vst1_u8_x3): Likewise.
16555         (vst1_u16_x3): Likewise.
16556         (vst1_u32_x3): Likewise.
16557         (vst1_f16_x3): Likewise.
16558         (vst1_f32_x3): Likewise.
16559         (vst1_p64_x3): Likewise.
16560         (vst1q_s8_x3): Likewise.
16561         (vst1q_p8_x3): Likewise.
16562         (vst1q_s16_x3): Likewise.
16563         (vst1q_p16_x3): Likewise.
16564         (vst1q_s32_x3): Likewise.
16565         (vst1q_s64_x3): Likewise.
16566         (vst1q_u8_x3): Likewise.
16567         (vst1q_u16_x3): Likewise.
16568         (vst1q_u32_x3): Likewise.
16569         (vst1q_u64_x3): Likewise.
16570         (vst1q_f16_x3): Likewise.
16571         (vst1q_f32_x3): Likewise.
16572         (vst1q_f64_x3): Likewise.
16573         (vst1q_p64_x3): Likewise.
16575 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16577         * config/msp430/msp430.c (msp430_output_labelref): Prepend
16578         user_label_prefix to name.
16580         * tree-core.h: Update comment about the format of NAME string
16581         passed to handler in attribute_spec.
16583         * config/msp430/msp430.md: Remove erroneous subreg expression from
16584         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
16585         zero_extend{q,h}isi2.
16587 2018-05-30  Borislav Petkov  <bp@suse.de>
16589         * doc/extend.texi: Document some architecture specific
16590         constraints and sort entries.
16592 2018-05-30  Martin Sebor  <msebor@redhat.com>
16594         PR middle-end/85369
16595         * builtins.c (expand_builtin_stpcpy_1): New function.
16596         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
16597         only if the former succeeds.
16599 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
16601         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
16602         in saphira.
16604 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
16606         * doc/invoke.texi (-flinker-output): Document
16608 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
16610         * passes.c (ipa_write_summaries): Only modify statements if body
16611         is in memory.
16612         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
16613         incrementally linking.
16614         (ipa_passes): Likewise.
16615         * lto-cgraph.c (lto_output_node): When incrementally linking do not
16616         pass down resolution info.
16617         * common.opt (flag_incremental_link): Update info.
16618         * gcc.c (plugin specs): Turn flinker-output=* to
16619         -plugin-opt=-linker-output-known
16620         * toplev.c (compile_file): Also cut compilation when doing incremental
16621         link.
16622         * flag-types. (enum lto_partition_model): Add
16623         LTO_LINKER_OUTPUT_NOLTOREL.
16624         (invoke.texi): Add -flinker-output docs.
16625         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
16626         link same way as WPA; do not stream in dead initializers.
16628         * dwarf2out.c (dwarf2out_die_ref_for_decl,
16629         darf2out_register_external_decl): Support incremental link.
16631 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
16633         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
16635 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
16637         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
16638         it down to simple_object_copy_lto_debug_sections.
16639         (run_gcc): Determine incremental LTO link time and configure
16640         lto1 into non-wpa mode, disable renaming of debug sections.
16642 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16644         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
16645         descriptions of various incorrectly documented functions.
16647 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16649         Revert:
16650         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
16651         address check not strict.
16653 2018-05-30  Richard Biener  <rguenther@suse.de>
16655         PR tree-optimization/85964
16656         * tracer.c (better_p): Drop initialized count check, we only
16657         call the function with initialized counts now.
16658         (find_best_successor): Do find a best edge if one
16659         has uninitialized count.
16660         (find_best_predecessor): Likewise.  Do BB frequency check only
16661         if count is initialized.
16663 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
16665         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
16666         (aarch64_ldrstr_offset_compare): New.
16667         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
16668         load/store orderings.
16669         (aarch64_gen_adjusted_ldpstp): Likewise.
16671 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
16673         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
16674         Check for subset of GENERAL_REGS and FP_REGS.
16675         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
16676         r=w alternative.
16678 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
16680         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
16681         and wi::to_poly_offset.  Add the current offset and then check
16682         whether the sum fits, rather than using an unchecked addition of
16683         a checked term.  Check for a shwi rather than a uhwi.
16684         * expr.c (get_bit_range): Use tree_to_poly_uint64.
16685         (store_constructor): Use poly_int_tree_p.
16686         (expand_expr_real_1): Likewise.
16687         * function.c (assign_temp): Likewise.
16688         * fold-const.c (const_binop): Use poly_int_tree_p and
16689         wi::to_poly_offset.
16690         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
16691         division.
16692         * ipa-icf-gimple.c (func_checker::compare_operand): Use
16693         to_poly_offset for MEM offsets.
16694         * ipa-icf.c (sem_variable::equals): Likewise.
16695         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
16696         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
16697         wi::to_poly_offset for BIT_FIELD_REF offsets.
16698         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
16699         wi::to_poly_offset.
16700         * var-tracking.c (emit_note_insn_var_location): Use
16701         tree_to_poly_uint64.
16703 2018-05-29  Jim Wilson  <jimw@sifive.com>
16705         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
16707 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
16709         PR target/85950
16710         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
16711         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
16712         sequence.
16713         (sse4_1_round<mode>2): Use nonimmediate_operand
16714         for operand 1 predicate.
16716 2018-05-29  Martin Sebor  <msebor@redhat.com>
16717             Richard Biener  <rguenther@suse.de>
16719         PR testsuite/85888
16720         * calls.c (get_size_range): Call determine_value_range instead
16721         of get_value_range..
16722         * tree-vrp.h (determine_value_range): Declared new function.
16723         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
16725 2018-05-29  Richard Biener  <rguenther@suse.de>
16727         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
16728         sure to use non-pattern stmts for get_earlier_stmt arguments.
16729         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
16730         called on pattern stmts.
16731         (get_later_stmt): Likewise.
16733 2018-05-29  Martin Liska  <mliska@suse.cz>
16735         PR gcov-profile/85759
16736         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
16737         env variables.
16739 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
16741         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
16742         VEC_UNPACK_*_EXPR.
16743         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
16744         VEC_PACK_*_EXPR.
16746         PR target/85918
16747         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
16748         VEC_PACK_FLOAT_EXPR): New tree codes.
16749         * tree-pretty-print.c (op_code_prio): Handle
16750         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
16751         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
16752         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
16753         * tree-inline.c (estimate_operator_cost): Likewise.
16754         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
16755         * fold-const.c (const_binop): Likewise.
16756         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
16757         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
16758         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
16759         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
16760         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
16761         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
16762         * expr.c (expand_expr_real_2): Likewise.
16763         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
16764         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
16765         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
16766         optabs.
16767         * optabs.c (expand_widen_pattern_expr): For
16768         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
16769         sign from result type rather than operand's type.
16770         (expand_binop_directly): For vec_packu_float_optab and
16771         vec_packs_float_optab allow result type to be different from operand's
16772         type.
16773         * optabs-tree.c (optab_for_tree_code): Handle
16774         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
16775         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
16776         * tree-vect-generic.c (expand_vector_operations_1):  Handle
16777         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
16778         VEC_PACK_FLOAT_EXPR.
16779         * tree-vect-stmts.c (supportable_widening_operation): Handle
16780         FIX_TRUNC_EXPR.
16781         (supportable_narrowing_operation): Handle FLOAT_EXPR.
16782         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
16783         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
16784         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
16785         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
16786         mode attributes.
16787         (vec_pack<floatprefix>_float_<mode>): New expander.
16788         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
16789         attributes.
16790         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
16791         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
16792         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
16793         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
16794         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
16795         Document.
16796         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
16797         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
16798         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
16799         VEC_PACK_FLOAT_EXPR): Document.
16801 2018-05-29  Richard Biener  <rguenther@suse.de>
16803         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
16804         member.
16805         (stmt_vec_info_vec): Make pointer.
16806         (init_stmt_vec_info_vec): Remove.
16807         (free_stmt_vec_info_vec): Likewise.
16808         (set_stmt_vec_info_vec): New function.
16809         (free_stmt_vec_infos): Likewise.
16810         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
16811         (set_vinfo_for_stmt): Likewise.
16812         (get_earlier_stmt): Likewise.
16813         (get_later_stmt): Likewise.
16814         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
16815         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
16816         (vec_info::~vec_info): Free stmt_vec_infos.
16817         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
16818         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
16819         (pass_slp_vectorize::execute): Likewise.
16820         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
16821         (free_stmt_vec_info_vec): Likewise.
16822         (set_stmt_vec_info_vec): New function.
16823         (free_stmt_vec_infos): Likewise.
16824         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
16825         the global stmt_vec_info_vec.
16826         * tree-parloops.c (gather_scalar_reductions): Use
16827         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
16828         vector.
16830 2018-05-29  Richard Biener  <rguenther@suse.de>
16832         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
16834 2018-05-29  Martin Liska  <mliska@suse.cz>
16835             David Malcolm  <dmalcolm@redhat.com>
16837         * vec.c (test_reverse): New.
16838         (vec_c_tests): Add new test.
16839         * vec.h (vl_ptr>::reverse): New function.
16841 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
16843         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
16845         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
16846         and later.
16848 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
16850         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
16852 2018-05-28  Richard Biener  <rguenther@suse.de>
16854         PR tree-optimization/85933
16855         * tree-vect-data-refs.c (vect_record_base_alignments): Only
16856         look at stmts marked as vectorizable.
16858 2018-05-28  Richard Biener  <rguenther@suse.de>
16860         PR tree-optimization/85934
16861         * tree-vect-generic.c (expand_vector_operations_1): Hoist
16862         vector boolean check before scalar optimization.
16864 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
16866         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
16867         for armv5te.
16869 2018-05-28  Mark Wielaard  <mark@klomp.org>
16871         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
16872         if it is an expression containing a minus sign.
16874 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
16876         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
16878 2018-05-27  Paul Koning  <ni1d@arrl.net>
16880         * config/pdp11/pdp11.md (truncsihi2): Remove.
16882 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
16883             Chung-Ju Wu  <jasonwucj@gmail.com>
16885         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
16886         implementation.
16887         (unaligned_store_dw): Ditto.
16888         * config/nds32/nds32-memory-manipulation.c
16889         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
16890         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
16891         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
16892         (emit_setmem_word_loop): Rename to ...
16893         (emit_setmem_doubleword_loop): ... this.
16894         (nds32_gen_dup_4_byte_to_word_value): New function.
16895         (nds32_gen_dup_8_byte_to_double_word_value): New function.
16896         (nds32_expand_setmem_loop): Refine implementation.
16897         (nds32_expand_setmem_loop_v3m): Ditto.
16898         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
16899         pattern.
16901 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
16903         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
16905 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
16907         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
16908         (nds32_init_machine_status): Initialize machine->attr_naked_p and
16909         machine->attr_no_prologue_p.
16910         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
16911         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
16912         (nds32_expand_epilogue): Consider attr_naked_p.
16913         (nds32_expand_epilogue_v3pop): Likewise.
16914         (nds32_can_use_return_insn): Likewise.
16915         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
16916         attr_no_prologue_p fields.
16917         * config/nds32/nds32.opt (mret-in-naked-func): New option.
16919 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
16921         PR target/85918
16922         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
16923         attributes.
16924         * config/i386/sse.md
16925         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
16926         Rename to ...
16927         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
16928         ... this.
16929         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
16930         Rename to ...
16931         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
16932         ... this.
16933         (*<floatsuffix>floatv2div2sf2): Rename to ...
16934         (*float<floatunssuffix>v2div2sf2): ... this.
16935         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
16936         (float<floatunssuffix>v2div2sf2_mask): ... this.
16937         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
16938         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
16939         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
16940         to ...
16941         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
16942         ... this.
16943         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
16944         Rename to ...
16945         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
16946         ... this.
16947         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
16948         Rename to ...
16949         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
16950         ... this.
16951         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
16952         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
16953         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
16954         gen_ufix_truncv8dfv8si2.
16955         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
16956         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
16957         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
16958         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
16959         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
16960         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
16961         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
16962         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
16964 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
16966         PR target/85900
16967         PR target/85345
16968         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
16970 2018-05-25  Jim Wilson  <jimw@sifive.com>
16972         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
16973         * config/riscv/riscv.c (struct machine_function): Add
16974         interrupt_handler_p and attribute_checked_p fields.
16975         (riscv_attribute_table): Add interrupt.
16976         (riscv_interrupt_type_p): New.
16977         (riscv_save_reg_p): Save extra regs for interrupt handler.
16978         (riscv_use_save_libcall): Return false  for interrupt handler.
16979         (riscv_first_stack_step): Add forward declaration.
16980         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
16981         for interrupt handler with large frame.  Use it for saved reg list.
16982         (riscv_expand_prologue): Move flag_stack_usage_info support to
16983         eliminate duplication.
16984         (riscv_expand_epilogue): Generate mret for interrupt handler.
16985         (riscv_epilogue_uses): New.
16986         (riscv_can_use_return_insn): Return false for interrupt handler.
16987         (riscv_function_ok_for_sibcall): Likewise.
16988         (riscv_set_current_function): Add interrupt handler support.
16989         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
16990         * config/riscv/riscv.md (UNSPECV_MRET): New.
16991         (GP_REGNUM): New.
16992         (riscv_frflags, riscv_fsflags): Use tab after opcode.
16993         (riscv_mret): New.
16994         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
16996 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
16998         PR tree-optimization/85712
16999         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
17000         this candidate has already been replaced in-situ by a copy.
17002 2018-05-25  Jason Merrill  <jason@redhat.com>
17004         PR c++/80485 - inline function non-zero address.
17005         * symtab.c (nonzero_address): Check DECL_COMDAT.
17007 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
17009         PR target/83628
17010         * config/alpha/alpha.md (ashlsi3): New insn pattern.
17011         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
17012         extension of SImode operation.  Use const123_operand predicate.
17013         (*saddsi_1): Remove.
17014         (*saddl_se_1): Ditto.
17015         (*ssubsi_1): Ditto.
17016         (*ssubl_se_1): Ditto.
17017         * config/alpha/predicates.md (const123_operand): New predicate.
17018         * config/alpha/constraints.md (P): Use IN_RANGE.
17020 2018-05-25  Richard Biener  <rguenther@suse.de>
17022         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
17023         defaulted to true.
17024         (ref_maybe_used_by_stmt_p): Likewise.
17025         (stmt_may_clobber_ref_p): Likewise.
17026         (stmt_may_clobber_ref_p_1): Likewise.
17027         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
17028         and pass it along.
17029         (ref_maybe_used_by_stmt_p): Likewise.
17030         (stmt_may_clobber_ref_p): Likewise.
17031         (stmt_may_clobber_ref_p_1): Likewise.
17032         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
17033         the alias oracle to disambiguate DRs with stmts DR analysis
17034         couldn't handle.
17035         (vect_analyze_data_refs): Do not give up on not analyzable
17036         DRs for BB vectorization.  Remove code truncating the dataref
17037         vector.
17039 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
17041         PR target/85832
17042         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17043         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
17044         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
17046 2018-05-25  Richard Biener  <rguenther@suse.de>
17048         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
17049         function, combining stmt data ref gathering and fatal analysis
17050         parts.
17051         (vect_analyze_data_refs): Remove now redudnant code and simplify.
17052         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
17053         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
17054         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
17055         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
17057 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
17059         PR tree-optimization/85720
17060         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
17061         SCC if all partitions are builtins.
17062         (version_loop_by_alias_check): New parameter.  Generate cancelable
17063         runtime alias check if all partitions are builtins.
17064         (distribute_loop): Update call to above function.
17066 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
17068         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
17069         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
17070         (parm_default_def_partition_arg): Ditto.
17071         (set_parm_default_def_partition): Ditto.
17072         (get_parm_default_def_partitions): Ditto and make it static.
17073         (get_undefined_value_partitions): Ditto and make it static.
17074         (remove_ssa_form): Refactor call to init_var_map here.
17075         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
17076         computation for loop region.
17077         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
17078         (register_default_def): Delete.
17079         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
17080         (parm_default_def_partition_arg): Ditto.
17081         (set_parm_default_def_partition): Ditto.
17082         (get_parm_default_def_partitions): Ditto and make it static.
17083         (get_undefined_value_partitions): Ditto and make it static.
17084         (coalesce_with_default, coalesce_with_default): Update comment.
17085         (create_coalesce_list_for_region): New func factored out from
17086         create_outofssa_var_map.
17087         (populate_coalesce_list_for_outofssa): New func factored out from
17088         create_outofssa_var_map and coalesce_ssa_name.
17089         (create_outofssa_var_map): Delete.
17090         (coalesce_ssa_name): Refactor to support live range computation.
17091         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
17092         (get_parm_default_def_partitions): Delete.
17093         (get_undefined_value_partitions): Ditto.
17094         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
17095         computation for loop region.
17096         (new_tree_live_info, loe_visit_block): Ditto.
17097         (live_worklist, set_var_live_on_entry): Ditto.
17098         (calculate_live_on_exit, verify_live_on_entry): Ditto.
17099         * tree-ssa-live.h (struct _var_map): New fields.
17100         (init_var_map): Change decl.
17101         (region_contains_p): New.
17103 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
17105         * tree-ssa-live.h (live_merge_and_clear): Delete.
17107 2018-05-25  Richard Biener  <rguenther@suse.de>
17109         PR c++/85912
17110         * tree-dump.c (dequeue_and_dump): Remove access to removed
17111         operand 2 of a SWITCH_EXPR.
17113 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17115         * doc/sourcebuild.texi (vect_double_cond_arith): Include
17116         multiplication and division.
17117         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
17118         (cond_udiv@var{m}, cond_umod@var{m}): Document.
17119         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
17120         (cond_udiv_optab, cond_umod_optab): New optabs.
17121         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
17122         (IFN_COND_RDIV): New internal functions.
17123         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
17124         TRUNC_MOD_EXPR and RDIV_EXPR.
17125         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
17126         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
17127         New unspecs.
17128         (SVE_INT_BINARY): Include mult.
17129         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
17130         (optab, sve_int_op): Handle mult.
17131         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
17132         UNSPEC_COND_DIV.
17133         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
17134         for SVE_INT_BINARY_SD.
17136 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17138         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
17139         (optab, sve_int_op): Handle div and udiv.
17140         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
17141         for SVE_INT_BINARY_SD.
17142         (*<optab><mode>3): New insn for the same.
17144 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17146         * tree-vect-patterns.c: Include predict.h.
17147         (vect_recog_divmod_pattern): Restrict check for division support
17148         to when optimizing for size.
17150 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17152         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
17153         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
17154         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
17155         (gimple_match_op::set_op): Likewise.
17156         (gimple_resimplify4): Declare.
17157         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
17158         (expr::gen_transform): Likewise.
17159         (decision_tree::gen): Generate a simplification routine for 4 operands.
17160         * gimple-match-head.c (gimple_simplify): Add an overload for
17161         4 operands.  In the top-level function, handle up to 4 call
17162         arguments and call gimple_resimplify4.
17163         (gimple_resimplify4): New function.
17164         (build_call_internal): Pass a fourth operand.
17165         (maybe_push_to_seq): Likewise.
17166         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
17167         Fold VEC_COND_EXPRs of an operation and a default value into
17168         an IFN_COND_* function if possible.
17169         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
17170         New unspecs.
17171         (SVE_COND_FP_BINARY): Include them.
17172         (optab, sve_fp_op): Handle them.
17173         (SVE_INT_BINARY_REV): New code iterator.
17174         (SVE_COND_FP_BINARY_REV): New int iterator.
17175         (commutative): New int attribute.
17176         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
17177         Declare.
17178         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
17179         function.
17180         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
17181         (*cond_<optab><mode>): New patterns for reversed operands.
17183 2018-05-25  Richard Biener  <rguenther@suse.de>
17185         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
17186         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
17187         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
17188         (STMT_VINFO_GROUPED_ACCESS): Adjust.
17189         * tree-vect-data-refs.c (everywhere): Adjust users.
17190         * tree-vect-loop.c (everywhere): Likewise.
17191         * tree-vect-slp.c (everywhere): Likewise.
17192         * tree-vect-stmts.c (everywhere): Likewise.
17193         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
17195 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17197         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
17198         Rename to...
17199         (gcc_cv_as_section_exclude): ... this.
17200         Try Solaris as #exclude syntax.
17201         * configure: Regenerate.
17202         * config.in: Regenerate.
17203         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
17204         SECTION_EXCLUDE.
17205         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
17206         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
17208         * varasm.c (default_elf_asm_named_section): Don't check if
17209         HAVE_GAS_SECTION_EXCLUDE is defined.
17211 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17213         * doc/md.texi: Update the documentation of the cond_* optabs
17214         to mention the new final operand.  Fix GET_MODE_NUNITS call.
17215         Describe the scalar case too.
17216         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
17217         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
17218         instead of 2.
17219         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
17220         (get_conditional_internal_fn): Update comment.
17221         * tree-vect-loop.c (vectorizable_reduction): Pass the original
17222         accumulator value as a final argument to conditional functions.
17223         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
17224         a define_expand and add an "else" operand.  Assert for now that
17225         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
17226         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
17227         (*cond_<optab><mode>): New patterns.
17228         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
17229         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
17230         (UNSPEC_COND_EOR): Delete.
17231         (optab): Remove associated mappings.
17232         (SVE_INT_BINARY): New code iterator.
17233         (sve_int_op): Remove int attribute and add "minus" to the code
17234         attribute.
17235         (SVE_COND_INT_OP): Delete.
17236         (SVE_COND_FP_OP): Rename to...
17237         (SVE_COND_FP_BINARY): ...this.
17239 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
17241         * optabs.c (can_reuse_operands_p): New function.
17242         (maybe_legitimize_operands): Try to reuse the results for
17243         earlier operands.
17245 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
17247         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
17248         Add {q} suffix to insn mnemonic.
17250 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17252         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
17253         (msp430_warn_func_return): New.
17255 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
17257         * fold-const.c (tree_nonzero_bits): New function.
17258         * fold-const.h (tree_nonzero_bits): Likewise.
17259         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
17260         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
17262 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
17264         PR target/85900
17265         PR target/85345
17266         * varasm.c (assemble_alias): Check ifunc_resolver only on
17267         FUNCTION_DECL.
17269 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
17271         PR target/85903
17272         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
17273         when memory input operand is handled.
17275 2018-05-24  Luis Machado  <luis.machado@linaro.org>
17277         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
17278         global.
17279         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
17281 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
17283         * match.pd: Delay FMA folds until after vectorization.
17285 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17287         PR target/83009
17288         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
17289         address check not strict.
17291 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
17293         * gimple-match.h (gimple_match_op): New class.
17294         (mprts_hook): Replace parameters with a gimple_match_op *.
17295         (maybe_build_generic_op): Likewise.
17296         (gimple_simplified_result_is_gimple_val): Replace parameters with
17297         a const gimple_match_op *.
17298         (gimple_simplify): Replace code_helper * and tree * parameters with
17299         a gimple_match_op * parameter.
17300         (gimple_resimplify1): Replace code_helper *, tree and tree *
17301         parameters with a gimple_match_op * parameter.
17302         (gimple_resimplify2): Likewise.
17303         (gimple_resimplify3): Likewise.
17304         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
17305         parameters with a gimple_match_op * parameter.
17306         * gimple-match-head.c (gimple_simplify): Change prototypes of
17307         auto-generated functions to take a gimple_match_op * instead of
17308         separate code_helper * and tree * parameters.  Make the same
17309         change in the top-level overload and update calls to the
17310         gimple_resimplify routines.  Update calls to the auto-generated
17311         functions and to maybe_push_res_to_seq in the publicly-facing
17312         operation-specific gimple_simplify overloads.
17313         (gimple_match_op::MAX_NUM_OPS): Define.
17314         (gimple_resimplify1): Replace rcode and ops with a single res_op
17315         parameter.  Update call to gimple_simplify.
17316         (gimple_resimplify2): Likewise.
17317         (gimple_resimplify3): Likewise.
17318         (mprts_hook): Replace parameters with a gimple_match_op *.
17319         (maybe_build_generic_op): Likewise.
17320         (build_call_internal): Replace type, nargs and ops with
17321         a gimple_match_op *.
17322         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
17323         with a single gimple_match_op *.  Update calls to mprts_hook,
17324         build_call_internal and gimple_simplified_result_is_gimple_val.
17325         Factor out code that is common to the tree_code and combined_fn cases.
17326         * genmatch.c (expr::gen_transform): Replace tem_code and
17327         tem_ops with a gimple_match_op called tem_op.  Update calls
17328         to the gimple_resimplify functions and maybe_push_res_to_seq.
17329         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
17330         res_ops.  Update call to the gimple_resimplify functions.
17331         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
17332         (decision_tree::gen): Make the functions take a gimple_match_op *
17333         called res_op instead of separate res_code and res_ops parameters.
17334         Update call accordingly.
17335         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
17336         and ops with a single res_op parameter.  Update calls to
17337         maybe_build_generic_op and maybe_push_res_to_seq.
17338         (fold_stmt_1): Update calls to gimple_simplify and
17339         replace_stmt_with_simplification.
17340         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
17341         and gimple_simplified_result_is_gimple_val.
17342         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
17343         gimple_simplify.
17344         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
17345         with a gimple_match_op *.
17346         (vn_nary_build_or_lookup): Likewise.  Update call to
17347         vn_nary_build_or_lookup_1.
17348         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
17349         gimple_match_op *.  Update calls to the gimple_resimplify routines
17350         and to gimple_simplified_result_is_gimple_val.
17351         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
17352         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
17353         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
17354         (visit_nary_op): Likewise.
17355         (visit_reference_op_load): Likewise.
17357 2018-05-23  Luis Machado  <luis.machado@linaro.org>
17359         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
17360         modifier for printing the step amount.
17362 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
17364         PR target/78849
17365         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
17366         types.
17368 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
17370         * doc/sourcebuild.texi (Endianness): New subsubsection.
17372 2018-05-23  Luis Machado  <luis.machado@linaro.org>
17374         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17375         <prefetch_dynamic_strides>: New const bool field.
17376         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17377         prefetch_dynamic_strides.
17378         (exynosm1_prefetch_tune): Likewise.
17379         (thunderxt88_prefetch_tune): Likewise.
17380         (thunderx_prefetch_tune): Likewise.
17381         (thunderx2t99_prefetch_tune): Likewise.
17382         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
17383         false.
17384         (aarch64_override_options_internal): Update to set
17385         PARAM_PREFETCH_DYNAMIC_STRIDES.
17386         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
17387         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
17388         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
17389         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
17390         prefetch-dynamic-strides setting.
17392 2018-05-23  Luis Machado  <luis.machado@linaro.org>
17394         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17395         <minimum_stride>: New const int field.
17396         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17397         minimum_stride field defaulting to -1.
17398         (exynosm1_prefetch_tune): Likewise.
17399         (thunderxt88_prefetch_tune): Likewise.
17400         (thunderx_prefetch_tune): Likewise.
17401         (thunderx2t99_prefetch_tune): Likewise.
17402         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
17403         <default_opt_level>: Set to 3.
17404         (aarch64_override_options_internal): Update to set
17405         PARAM_PREFETCH_MINIMUM_STRIDE.
17406         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
17407         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
17408         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
17409         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
17410         stride is constant and is below the minimum stride threshold.
17412 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17414         * config/arm/arm-cpus.in (mode26): Delete.
17415         (armv4): Delete mode26 reference.
17416         * config/arm/arm.c (arm_configure_build_target): Delete use of
17417         isa_bit_mode26.
17419 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
17421         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
17422         New insn pattern.
17423         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
17424         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
17425         for non-SSE modes.
17426         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
17427         (floatunsdidf2): Ditto.
17429 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
17431         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
17432         (fixuns_trunc<mode>si2_avx512f): Ditto.
17433         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
17434         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
17435         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
17437 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
17439         PR rtl-optimization/79985
17440         * df-scan.c (df_insn_refs_collect): Remove special case for
17441         global registers and asm statements.
17443 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
17445         * extend.texi (Global Register Variables): Rewrite the bullet list.
17446         Note that the register is available for allocation. Note that access
17447         via inline asm must use constraints. Add note about async-signal
17448         handlers. Remove paragraph about automagic register selection.
17450 2018-05-23  Richard Biener  <rguenther@suse.de>
17452         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
17453         of fixed offset from memset VN.
17455 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
17457         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
17458         first_interp field.
17459         (alloc_cand_and_find_basis): Initialize first_interp field.
17460         (slsr_process_mul): Modify first_interp field.
17461         (slsr_process_add): Likewise.
17462         (slsr_process_cast): Modify first_interp field for each new
17463         interpretation.
17464         (slsr_process_copy): Likewise.
17465         (dump_candidate): Dump first_interp field.
17466         (replace_mult_candidate): Process all interpretations, not just
17467         subsequent ones.
17468         (replace_rhs_if_not_dup): Likewise.
17469         (replace_one_candidate): Likewise.
17471 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
17473         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
17474         Add new boolean.
17475         (aarch64_needs_frame_chain): New function.
17476         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
17478 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
17480         PR target/84882
17481         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
17482         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
17483         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
17484         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
17485         as true for strict-align.
17486         (aarch64_can_inline_p): Perform checks even when callee has no
17487         attributes to check for strict alignment.
17488         * doc/extend.texi (AArch64 Function Attributes): Document
17489         no-strict-align.
17490         * doc/invoke.texi: (AArch64 Options): Likewise.
17492 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
17494         PR tree-optimization/85853
17495         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
17496         the handling of the root of the node to...
17497         (vect_slp_analyze_node_operations_1): ...this new function,
17498         and run the whole thing with the child nodes' def types
17499         set according to their SLP node's def type.
17501 2018-05-23  Richard Biener  <rguenther@suse.de>
17503         PR middle-end/85874
17504         * tree-data-ref.c (create_runtime_alias_checks): Defer
17505         and ignore overflow warnings.
17507 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
17509         PR tree-optimization/85822
17510         * tree-vrp.c (is_masked_range_test): Fix handling of negative
17511         constants.
17513 2018-05-23  Richard Biener  <rguenther@suse.de>
17515         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
17516         memset constants via native_interpret_expr.
17518 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
17520         PR target/85345
17521         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
17522         attribute.
17523         (cgraph_node::create_alias): Likewise.
17524         (cgraph_node::get_availability): Check ifunc_resolver instead
17525         of looking up ifunc attribute.
17526         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
17527         * varasm.c (do_assemble_alias): Likewise.
17528         (assemble_alias): Likewise.
17529         (default_binds_local_p_3): Likewise.
17530         * cgraph.h (cgraph_node): Add ifunc_resolver.
17531         (cgraph_node::only_called_directly_or_aliased_p): Return false
17532         for IFUNC resolver.
17533         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
17534         attribute.
17535         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
17536         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
17537         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
17538         instead of looking up ifunc attribute.
17540 2018-05-22  Luis Machado  <luis.machado@linaro.org>
17542         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
17544 2018-05-22  Martin Sebor  <msebor@redhat.com>
17546         PR middle-end/85359
17547         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
17548         only when expasion succeeds.
17549         (expand_builtin_strcmp): Same.
17550         (expand_builtin_strncmp): Same.
17552 2018-05-22  Martin Sebor  <msebor@redhat.com>
17554         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
17556 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
17557             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17559         * config/aarch64/aarch64-ldpstp.md: Replace uses of
17560         aarch64_mem_pair_operand with memory_operand and delete operand swapping
17561         code.
17562         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
17563         Add check for legitimate_address.
17564         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
17565         (aarch64_swap_ldrstr_operands): New.
17566         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
17567         Define prototype.
17569 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
17570             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17572         * config/aarch64/aarch64.md: New patterns to generate stp
17573         and ldp.
17574         (store_pair_sw, store_pair_dw): New patterns to generate stp for
17575         single words and double words.
17576         (load_pair_sw, load_pair_dw): Likewise.
17577         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
17578         Delete.
17579         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
17580         Delete.
17581         * config/aarch64/aarch64-ldpstp.md: Modify peephole
17582         for different mode ldpstp and add peephole for merged zero stores.
17583         Likewise for loads.
17584         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
17585         Add size check.
17586         (aarch64_gen_store_pair): Rename calls to match new patterns.
17587         (aarch64_gen_load_pair): Rename calls to match new patterns.
17588         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
17589         (load_pair<DREG:mode><DREG2:mode>): ... This.
17590         (store_pair<mode>): Rename to...
17591         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
17592         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
17593         New mode iterators.
17594         (V_INT_EQUIV): Handle SImode.
17595         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
17596         New predicate.
17598 2018-05-22  Martin Sebor  <msebor@redhat.com>
17600         PR c/85623
17601         * calls.c (maybe_warn_nonstring_arg): Use string length to set
17602         or ajust the presumed bound on an operation to avoid unnecessary
17603         warnings.
17605 2018-05-22  Martin Sebor  <msebor@redhat.com>
17607         PR tree-optimization/85826
17608         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
17609         assuming that a DECL necesarily has a constant size.
17611 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
17613         PR middle-end/85862
17614         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
17616 2018-05-22  Richard Biener  <rguenther@suse.de>
17618         PR tree-optimization/85834
17619         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
17620         non-constant and non-zero memset arguments.
17622 2018-05-22  Martin Liska  <mliska@suse.cz>
17624         PR ipa/85607
17625         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
17627 2018-05-22  Richard Biener  <rguenther@suse.de>
17629         PR tree-optimization/85863
17630         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
17631         comparisons when vectype is specified.
17632         (vectorizable_condition): Do not specify vectype for
17633         vect_is_simple_cond when SLP vectorizing.
17635 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
17637         PR target/85657
17638         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
17639         define __ibm128 as long double.
17640         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
17641         as a distinct type when IEEE 128-bit support is enabled.
17642         (init_float128_ieee): Fix up conversions between IFmode and IEEE
17643         128-bit types to use the correct functions.
17644         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
17645         convert between 128-bit floating point types that have different
17646         modes but the same representation, instead of using gen_lowpart to
17647         makean alias.
17648         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
17649         KFmode.
17650         (IFKF_reg): New attributes to give the register constraints for
17651         IFmode and KFmode.
17652         (extend<mode>tf2_internal): New insns to mark an explicit
17653         conversion between 128-bit floating point types that have a
17654         different mode but share the same representation.
17656 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
17658         PR tree-optimization/85814
17659         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
17660         a null return from get_strinfo when unsharing the next
17661         strinfo in the chain.
17663 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
17665         PR gcc/84923
17666         * varasm.c (weak_finish): Clean up weak_decls.
17668 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17670         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
17671         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
17672         UNSPEC_UADALP values.
17673         * config/aarch64/iterators.md (ABAL): New int iterator.
17674         (ABDL2): Likewise.
17675         (ADALP): Likewise.
17676         (sur): Add mappings for the above.
17677         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
17678         New define_insn.
17679         (aarch64_<sur>abal<mode>_4): Likewise.
17680         (aarch64_<sur>adalp<mode>_3): Likewise.
17681         (<sur>sadv16qi): New define_expand.
17683 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
17685         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
17686         (*movdf_internal): Ditto.
17687         (*rcpsf2_sse): Ditto.
17688         (*rsqrtsf2_sse): Ditto.
17689         (*sqrt<mode>2_sse): Ditto.
17691 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
17693         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
17694         eor3q<mode>4.
17695         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
17696         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
17697         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
17698         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
17699         vbcaxq_s64): New.
17700         * config/aarch64/arm_neon.h: Likewise.
17701         * config/aarch64/iterators.md (VQ_I): New.
17703 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
17705         * config.gcc: Add arc/t-multilib-linux to tmake_file for
17706         arc*-*-linux*.
17707         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
17708         MULTILIB_DIRNAMES
17710 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
17712         * config/nds32/constraints.md (S): New constraint.
17713         * config/nds32/nds32.md (call_internal): Use constraint S.
17714         (call_value_internal): Likewise.
17715         (sibcall_internal): Likewise.
17716         (sibcall_value_internal): Likewise.
17718 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
17719             Chung-Ju Wu  <jasonwucj@gmail.com>
17721         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
17722         into consideration.
17724 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
17725             Chung-Ju Wu  <jasonwucj@gmail.com>
17727         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
17728         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
17729         (nds32_rtx_costs_impl): Simplify.
17730         (nds32_address_cost_impl): Simplify.
17731         (nds32_init_rtx_costs): New function.
17732         (nds32_rtx_costs_speed_prefer): Likewise.
17733         (nds32_rtx_costs_size_prefer): Likewise.
17734         (nds32_address_cost_speed_prefer): Likewise.
17735         (nds32_address_cost_speed_fwprop): Likewise.
17736         (nds32_address_cost_size_prefer): Likewise.
17737         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
17738         * config/nds32/nds32.c (nds32_option_override): Use
17739         nds32_init_rtx_costs function.
17741 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
17743         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
17744         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
17745         (TARGET_PIPELINE_N8): Likewise.
17746         (TARGET_PIPELINE_N10): Likewise.
17747         (TARGET_PIPELINE_N13): Likewise.
17748         (TARGET_PIPELINE_GRAYWOLF): Likewise.
17750 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
17752         * config/nds32/nds32-fpu.md: Update copyright year.
17754 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
17756         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
17758 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
17760         * config/nds32/nds32.c
17761         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
17762         * config/nds32/nds32.opt (minline-asm-r15): New option.
17764 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
17766         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
17767         MASK_HW_ABS.
17768         * config/nds32/nds32.md (abssi2): New pattern.
17770 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
17772         * config/i386/i386.md (rex64namesuffix): New mode attribute.
17773         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
17774         Merge insn pattern from sse_cvtsi2ss<round_name> and
17775         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
17776         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
17777         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
17778         using SWI48 mode iterator.
17779         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
17780         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
17781         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
17782         pattern from sse_cvttss2si<round_saeonly_name>
17783         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
17784         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
17785         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
17786         using SWI48 mode iterator.
17787         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
17788         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
17789         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
17790         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
17791         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
17792         using SWI48 mode iterator.
17793         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
17794         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
17795         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
17796         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
17797         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
17798         SWI48 mode iterator.
17799         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
17800         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
17801         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
17802         pattern from sse_cvttsd2si<round_saeonly_name>
17803         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
17805 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
17807         * config/nds32/nds32-md-auxiliary.c
17808         (nds32_valid_smw_lwm_base_p): Refine.
17809         (nds32_output_smw_single_word): Refine.
17810         (nds32_output_smw_double_word): New.
17811         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
17813 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
17815         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
17816         (nds32_output_stack_pop): Refine.
17817         (nds32_expand_unaligned_load): Refine.
17818         (nds32_expand_unaligned_store): Refine.
17820 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
17821             Chung-Ju Wu  <jasonwucj@gmail.com>
17823         * config/nds32/constants.md: Add TP_REGNUM constant.
17824         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
17825         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
17826         UNSPEC_ADD32.
17827         * config/nds32/nds32-doubleword.md: Consider flag_pic.
17828         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
17829         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
17830         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
17831         and PIC code generation.
17832         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
17833         code generation.
17834         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
17835         optimization.
17836         * config/nds32/nds32.md: Support TLS and PIC.
17837         * config/nds32/nds32.c: Support TLS and PIC.
17838         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
17839         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
17840         predicate.
17842 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
17844         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
17845         mode with E_ prefix.
17847 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
17848             Chung-Ju Wu  <jasonwucj@gmail.com>
17850         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
17851         * config/nds32/nds32-md-auxiliary.c
17852         (symbolic_reference_mentioned_p): New.
17853         (nds32_legitimize_ict_address): New.
17854         (nds32_expand_ict_move): New.
17855         (nds32_indirect_call_referenced_p): New.
17856         (nds32_symbol_binds_local_p): Delete.
17857         (nds32_long_call_p): Modify.
17858         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
17859         * config/nds32/nds32-protos.h
17860         (symbolic_reference_mentioned_p): Declare.
17861         (nds32_legitimize_ict_address): Declare.
17862         (nds32_expand_ict_move): Declare.
17863         (nds32_indirect_call_referenced_p): Declare.
17864         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
17865         (nds32_relax_group): Use nds32_ict_const_p as condition.
17866         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
17867         (nds32_asm_file_start): Output ict_model directive in asm code.
17868         (nds32_legitimate_address_p): Consider indirect call.
17869         (nds32_print_operand): Consider indirect call.
17870         (nds32_print_operand_address): Consider indirect call.
17871         (nds32_insert_attributes): Handle "indirect_call" attribute.
17872         (TARGET_LEGITIMATE_ADDRESS_P): Define.
17873         (TARGET_LEGITIMATE_CONSTANT_P): Define.
17874         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
17875         (TARGET_DELEGITIMIZE_ADDRESS): Define.
17876         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
17877         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
17878         (TARGET_ICT_MODEL_SMALL): Define.
17879         (TARGET_ICT_MODEL_LARGE): Define.
17880         * config/nds32/nds32.md (movsi): Consider ict model.
17881         (call, call_value): Consider ict model.
17882         (sibcall, sibcall_value): Consider ict model.
17883         * config/nds32/nds32.opt (mict-model): New option.
17884         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
17885         model.
17887 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
17888             Monk Chiang  <sh.chiang04@gmail.com>
17889             Jim Wilson <jimw@sifive.com>
17891         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
17892         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
17893         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
17894         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
17895         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
17896         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
17897         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
17898         compute save_libcall_adjustment properly.
17899         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
17900         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
17901         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
17902         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
17903         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
17904         (ABI_SPEC): Handle mabi=ilp32e.
17905         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
17906         (RVE): Add RVE mask.
17907         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
17908         <-march>: Add rv32e as an example.
17910 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
17912         PR c++/82899
17913         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
17914         (intra_create_variable_infos): Handle C++ constructors.
17916 2018-05-18  Martin Liska  <mliska@suse.cz>
17918         * passes.def: Remove a redundant pass.
17920 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
17922         PR bootstrap/85838
17923         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
17925 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17927         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
17928         (ARMv4): Update.
17929         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
17930         (ARMv6m): Update.
17931         (armv2, armv2a, armv3, armv3m): Delete architectures.
17932         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
17933         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
17934         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
17935         Delete cpus.
17936         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
17937         (*mulsidi3adddi): Likewise.
17938         (mulsidi3): Likewise.
17939         (*mulsidi3_nov6): Likewise.
17940         (umulsidi3): Likewise.
17941         (umulsidi3_nov6): Likewise.
17942         (umaddsidi4): Likewise.
17943         (*umulsidi3adddi): Likewise.
17944         (smulsi3_highpart): Likewise.
17945         (*smulsi3_highpart_nov6): Likewise.
17946         (umulsi3_highpart): Likewise.
17947         (*umulsi3_highpart_nov6): Likewise.
17948         * config/arm/arm.h (arm_arch3m): Delete.
17949         * config/arm/arm.c (arm_arch3m): Delete.
17950         (arm_option_override_internal): Update armv3-related comment.
17951         (arm_configure_build_target): Delete use of isa_bit_mode32.
17952         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
17953         (arm_rtx_costs_internal): Delete check of arm_arch3m.
17954         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
17955         (mulsa3): Likewise.
17956         (mulusa3): Likewise.
17957         * config/arm/arm-protos.h (arm_arch3m): Delete.
17958         * config/arm/arm-tables.opt: Regenerate.
17959         * config/arm/arm-tune.md: Likewise.
17960         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
17961         deleted architectures.
17963 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17965         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
17966         (armv5t, armv5te): New features.
17967         (ARMv5, ARMv5e): Delete fgroups.
17968         (ARMv5t, ARMv5te): Adjust for above changes.
17969         (ARMv6m): Likewise.
17970         (armv5, armv5e): Delete arches.
17971         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
17972         arm_arch5.
17973         (*call_reg_arm): Likewise.
17974         (*call_value_reg_armv5): Likewise.
17975         (*call_value_reg_arm): Likewise.
17976         (*call_symbol): Likewise.
17977         (*call_value_symbol): Likewise.
17978         (*sibcall_insn): Likewise.
17979         (*sibcall_value_insn): Likewise.
17980         (clzsi2): Likewise.
17981         (prefetch): Likewise.
17982         (define_split and define_peephole2 dependent on arm_arch5):
17983         Likewise.
17984         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
17985         arm_arch5e.
17986         (TARGET_ARM_QBIT): Likewise.
17987         (TARGET_DSP_MULTIPLY): Likewise.
17988         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
17989         (arm_arch5, arm_arch5e): Delete.
17990         (arm_arch5t, arm_arch5te): Declare.
17991         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
17992         (arm_arch5t): Declare.
17993         (arm_option_reconfigure_globals): Update for the above.
17994         (arm_options_perform_arch_sanity_checks): Update comment, replace
17995         use of arm_arch5 with arm_arch5t.
17996         (use_return_insn): Likewise.
17997         (arm_emit_call_insn): Likewise.
17998         (output_return_instruction): Likewise.
17999         (arm_final_prescan_insn): Likewise.
18000         (arm_coproc_builtin_available): Likewise.
18001         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
18002         arm_arch5e with arm_arch5t and arm_arch5te.
18003         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
18004         (arm_arch5t, arm_arch5te): Declare.
18005         * config/arm/arm-tables.opt: Regenerate.
18006         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
18007         * config/arm/t-multilib: Likewise.
18008         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
18009         instead of arm_arch5.
18010         (*call_reg_thumb1): Likewise.
18011         (*call_value_reg_thumb1_v5): Likewise.
18012         (*call_value_reg_thumb1): Likewise.
18013         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
18014         unreachable path.
18015         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
18017 2018-05-18  Martin Liska  <mliska@suse.cz>
18019         PR gcov-profile/84846
18020         * doc/gcov.texi: Document -t option of gcov tool.
18022 2018-05-18  Martin Liska  <mliska@suse.cz>
18024         PR gcov-profile/84846
18025         * gcov.c (print_usage): Add new -t option.
18026         (process_args): Handle the option.
18027         (generate_results): Use stdout as output when requested by
18028         the option.
18030 2018-05-18  Martin Liska  <mliska@suse.cz>
18032         PR gcov-profile/84846
18033         * coverage.c (coverage_init): Write PWD to .gcno file.
18034         * doc/gcov.texi: Document how working directory is printed.
18035         * gcov-dump.c (dump_gcov_file): Print PWD.
18036         * gcov.c (output_intermediate_file): Likewise.
18037         (read_graph_file): Read PWD string.
18038         (output_lines): Print PWD.
18040 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18042         PR middle-end/85817
18043         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
18044         for retval and return false if all args to phi are zero.
18046 2018-05-18  Richard Biener  <rguenther@suse.de>
18048         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
18049         method.
18050         (evrp_dom_walker::before_dom_children): Call it.
18052 2018-05-18  Richard Biener  <rguenther@suse.de>
18054         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
18055         results when processing array refs with variable index.
18057 2018-05-18  Toon Moene  <toon@moene.org>
18059         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
18060         directly after that of -floop-interchange. Indicate that both
18061         options are enabled by default when specifying -O3.
18063 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18065         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
18066         iterator.  Delete separate integer-mode vec_set<mode> expander.
18067         (aarch64_simd_vec_setv2di): Delete.
18068         (vec_setv2di): Delete.
18069         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
18070         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
18071         the "w, r" alternative.
18073 2018-05-18  Martin Liska  <mliska@suse.cz>
18075         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
18076         * tree-pass.h (make_pass_lower_switch_O0): New function.
18077         * tree-switch-conversion.c (node_has_low_bound): Remove.
18078         (node_has_high_bound): Likewise.
18079         (node_is_bounded): Likewise.
18080         (class pass_lower_switch): Make it a template type and create
18081         two instances.
18082         (pass_lower_switch::execute): Add template argument.
18083         (make_pass_lower_switch): New function.
18084         (make_pass_lower_switch_O0): New function.
18085         (do_jump_if_equal): Remove.
18086         (emit_case_nodes): Simplify to just handle all 3 cases and leave
18087         all the hard work to tree optimization passes.
18089 2018-05-18  Martin Liska  <mliska@suse.cz>
18091         * dbgcnt.c (limit_low): Renamed from limit.
18092         (limit_high): New variable.
18093         (dbg_cnt_is_enabled): Check for upper limit.
18094         (dbg_cnt): Adjust dumping.
18095         (dbg_cnt_set_limit_by_index): Add new argument for high
18096         value.
18097         (dbg_cnt_set_limit_by_name): Likewise.
18098         (dbg_cnt_process_single_pair): Parse new format.
18099         (dbg_cnt_process_opt): Use strtok.
18100         (dbg_cnt_list_all_counters): Remove 'value' and add
18101         'limit_high'.
18102         * doc/invoke.texi: Document changes.
18104 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
18106         * doc/sourcebuild.texi (scalar_all_fma): Document.
18107         * tree.def (FMA_EXPR): Delete.
18108         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
18109         * internal-fn.c (ternary_direct): New macro.
18110         (expand_ternary_optab_fn): Likewise.
18111         (direct_ternary_optab_supported_p): Likewise.
18112         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
18113         * builtins.c (fold_builtin_fma): Delete.
18114         (fold_builtin_3): Don't call it.
18115         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
18116         * expr.c (expand_expr_real_2): Likewise.
18117         * fold-const.c (operand_equal_p): Likewise.
18118         (fold_ternary_loc): Likewise.
18119         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
18120         * gimple.c (DEFTREECODE): Likewise.
18121         * gimplify.c (gimplify_expr): Likewise.
18122         * optabs-tree.c (optab_for_tree_code): Likewise.
18123         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18124         * tree-eh.c (operation_could_trap_p): Likewise.
18125         (stmt_could_throw_1_p): Likewise.
18126         * tree-inline.c (estimate_operator_cost): Likewise.
18127         * tree-pretty-print.c (dump_generic_node): Likewise.
18128         (op_code_prio): Likewise.
18129         * tree-ssa-loop-im.c (stmt_cost): Likewise.
18130         * tree-ssa-operands.c (get_expr_operands): Likewise.
18131         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
18132         * fold-const-call.h (fold_fma): Delete.
18133         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
18134         CFN_FNMA and CFN_FNMS.
18135         (fold_fma): Delete.
18136         * genmatch.c (combined_fn): New enum.
18137         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
18138         (commutative_op): New function.
18139         (commutate): Use it.  Handle more than 2 operands.
18140         (dt_operand::gen_gimple_expr): Use commutative_op.
18141         (parser::parse_expr): Allow :c to be used with non-binary
18142         operators if the commutative operand is known.
18143         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
18144         CFN_FMS, CFN_FNMA and CFN_FNMS.
18145         (backprop::process_assign_use): Remove FMA_EXPR handling.
18146         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
18147         (gen_hsa_fma): New function.
18148         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
18149         IFN_FNMA and IFN_FNMS.
18150         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
18151         * gimple-fold.h (follow_all_ssa_edges): Declare.
18152         * gimple-fold.c (follow_all_ssa_edges): New function.
18153         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
18154         gimple_build interface and use follow_all_ssa_edges to fold the result.
18155         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
18156         instead of checking for optabs directly.
18157         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
18158         rather than FMA_EXPRs.
18159         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
18160         call to IFN_FMA instead of an FMA_EXPR.
18162 2018-05-17  Jim Wilson  <jimw@sifive.com>
18164         * expr.c (do_tablejump): When converting index to Pmode, if we have a
18165         sign extended promoted subreg, and the range does not have the sign bit
18166         set, then do a sign extend.
18168         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
18169         test, check for sign extended subreg and/or constant operands, and
18170         do a sign extend in that case.
18172 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
18174         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
18175         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
18176         Add untyped.
18177         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
18178         Change logics_shift_reg to logics_shift_imm.
18179         (thunderx2t99_fp_loadpair_basic): Delete.
18180         (thunderx2t99_fp_storepair_basic): Delete.
18181         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
18182         (thunderx2t99_asimd_polynomial): Delete.
18183         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
18184         and neon_fp_mul_d_scalar_q.
18185         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
18186         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
18187         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
18188         (thunderx2t99_asimd_lut): Add missing tbl types.
18189         (thunderx2t99_asimd_ext): Delete.
18190         (thunderx2t99_asimd_load1_1_mult): Delete.
18191         (thunderx2t99_asimd_load1_2_mult): Delete.
18192         (thunderx2t99_asimd_load1_ldp): New.
18193         (thunderx2t99_asimd_load1): New.
18194         (thunderx2t99_asimd_load2): Add missing *load2* types.
18195         (thunderx2t99_asimd_load3): New.
18196         (thunderx2t99_asimd_load4): New.
18197         (thunderx2t99_asimd_store1_1_mult): Delete.
18198         (thunderx2t99_asimd_store1_2_mult): Delete.
18199         (thunderx2t99_asimd_store2_mult): Delete.
18200         (thunderx2t99_asimd_store2_onelane): Delete.
18201         (thunderx2t99_asimd_store_stp): New.
18202         (thunderx2t99_asimd_store1): New.
18203         (thunderx2t99_asimd_store2): New.
18204         (thunderx2t99_asimd_store3): New.
18205         (thunderx2t99_asimd_store4): New.
18207 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
18209         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
18210         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
18212 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
18213             Segher Boessenkool  <segher@kernel.crashing.org>
18215         PR target/85698
18216         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
18217         operand.
18219 2018-05-17  Richard Biener  <rguenther@suse.de>
18221         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
18222         for pruning loop and prune defs feeding only already visited PHIs.
18224 2018-05-17  Richard Biener  <rguenther@suse.de>
18226         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
18228 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
18229             Richard Biener  <rguenther@suse.de>
18231         PR tree-optimization/85793
18232         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
18233         for VMAT_ELEMENTWISE.
18235 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
18237         * internal-fn.h (lookup_internal_fn): Declare
18238         * internal-fn.c (lookup_internal_fn): New function.
18239         * gimple.c (gimple_build_call_from_tree): Handle calls to
18240         internal functions.
18241         * gimple-pretty-print.c (dump_gimple_call): Print "." before
18242         internal function names.
18243         * tree-pretty-print.c (dump_generic_node): Likewise.
18244         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
18246 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
18248         * gimple-fold.h (gimple_build): Make the function forms take
18249         combined_fn rather than built_in_function.
18250         (gimple_simplify): Likewise.
18251         * gimple-match-head.c (gimple_simplify): Likewise.
18252         * gimple-fold.c (gimple_build): Likewise.
18253         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
18254         rather than gimple_build_call_internal.
18255         (get_initial_defs_for_reduction): Likewise.
18256         (vect_create_epilog_for_reduction): Likewise.
18257         (vectorizable_live_operation): Likewise.
18259 2018-05-17  Martin Liska  <mliska@suse.cz>
18261         * gimple-ssa-sprintf.c (format_directive): Do not use
18262         space in between 'G_' and '('.
18264 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
18266         PR target/85323
18267         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
18268         even if the mask is not all ones.
18270         PR target/85323
18271         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
18272         vector.
18273         (ix86_gimple_fold_builtin): Likewise.
18275         PR target/85323
18276         * config/i386/i386.c: Include tree-vector-builder.h.
18277         (ix86_vector_shift_count): New function.
18278         (ix86_fold_builtin): Fold shift builtins by scalar count.
18279         (ix86_gimple_fold_builtin): Likewise.
18281         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
18282         _mm512_setzero): New intrinsics.
18284 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
18285             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18287         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
18288         code generation for cases where splatting a value is not useful.
18289         * simplify-rtx.c (simplify_ternary_operation): Simplify
18290         vec_merge across a vec_duplicate and a paradoxical subreg forming
18291         a vector mode to a vec_concat.
18293 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
18295         * config.gcc: Support "goldmont-plus".
18296         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
18297         "goldmont-plus".
18298         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18299         PROCESSOR_GOLDMONT_PLUS.
18300         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
18301         (processor_target_table): Add "goldmont-plus".
18302         (PTA_GOLDMONT_PLUS): Define.
18303         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
18304         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
18305         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
18306         (fold_builtin_cpu): Add "goldmont-plus".
18307         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
18308         (ix86_option_override_internal): Add "goldmont-plus".
18309         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
18310         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
18311         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
18312         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
18314 2018-05-17  Richard Biener  <rguenther@suse.de>
18316         PR tree-optimization/85757
18317         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
18318         remove defs that only feed that PHI from further processing.
18320 2018-05-16  Jim Wilson  <jimw@sifive.com>
18322         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
18323         asterisk to name.
18324         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
18326 2018-05-16  Mark Wielaard  <mark@klomp.org>
18328         * dwarf2out.c (count_index_strings): New function.
18329         (output_indirect_strings): Call count_index_strings and generate
18330         header for dwarf_version >= 5.
18332 2018-05-16  Mark Wielaard  <mark@klomp.org>
18334         * dwarf2out.c (dwarf_FORM): New function.
18335         (set_indirect_string): Use dwarf_FORM.
18336         (reset_indirect_string): Likewise.
18337         (size_of_die): Likewise.
18338         (value_format): Likewise.
18339         (output_die): Likewise.
18340         (add_skeleton_AT_string): Likewise.
18341         (output_macinfo_op): Likewise.
18342         (index_string): Likewise.
18343         (output_index_string_offset): Likewise.
18344         (output_index_string): Likewise.
18345         (count_index_strings): Likewise.
18347 2018-05-16  Carl Love  <cel@us.ibm.com>
18349         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
18350         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
18352 2018-05-16  Martin Jambor  <mjambor@suse.cz>
18354         * ipa-prop.c (ipa_free_all_edge_args): Remove.
18355         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
18357 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
18359         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
18360         (fnma<mode>4): Likewise.
18361         (fms<mode>4): Likewise.
18362         (fnms<mode>4): Likewise.
18363         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
18364         (aarch64_fnma<mode>4): Likewise.
18365         (aarch64_fms<mode>4): Likewise.
18366         (aarch64_fnms<mode>4): Likewise.
18367         (aarch64_fnmadd<mode>4): Likewise.
18369 2018-05-16  Jason Merrill  <jason@redhat.com>
18371         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
18373 2018-05-16  Richard Biener  <rguenther@suse.de>
18375         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
18376         (dump_stmt_cost): Declare.
18377         (add_stmt_cost): Dump cost we add.
18378         (add_stmt_costs): New function.
18379         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
18380         No longer exported.
18381         (vect_analyze_stmt): Adjust prototype.
18382         (vectorizable_condition): Likewise.
18383         (vectorizable_live_operation): Likewise.
18384         (vectorizable_reduction): Likewise.
18385         (vectorizable_induction): Likewise.
18386         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
18387         cost vector to pass to vectorizable_ and record afterwards.
18388         (vect_model_reduction_cost): Take cost vector argument and adjust.
18389         (vect_model_induction_cost): Likewise.
18390         (vectorizable_reduction): Likewise.
18391         (vectorizable_induction): Likewise.
18392         (vectorizable_live_operation): Likewise.
18393         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
18394         SLP_TREE_NUMBER_OF_VEC_STMTS.
18395         (vect_analyze_slp_cost_1): Remove.
18396         (vect_analyze_slp_cost): Likewise.
18397         (vect_slp_analyze_node_operations): Take visited args and
18398         a target cost vector.  Avoid processing already visited stmt sets.
18399         (vect_slp_analyze_operations): Use a local cost vector to gather
18400         costs and register those of non-discarded instances.
18401         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
18402         (vect_schedule_slp_instance): Remove copying of
18403         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
18404         zero.
18405         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
18406         adding cost.  Record cost entry location.
18407         (vect_prologue_cost_for_slp_op): Function to compute cost of
18408         a constant or invariant generated for SLP vect in the prologue,
18409         split out from vect_analyze_slp_cost_1.
18410         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
18411         (vect_model_promotion_demotion_cost): Likewise.
18412         (vect_model_store_cost): Likewise, make static.
18413         (vect_model_load_cost): Likewise.
18414         (vectorizable_bswap): Add cost vector arg and adjust.
18415         (vectorizable_call): Likewise.
18416         (vectorizable_simd_clone_call): Likewise.
18417         (vectorizable_conversion): Likewise.
18418         (vectorizable_assignment): Likewise.
18419         (vectorizable_shift): Likewise.
18420         (vectorizable_operation): Likewise.
18421         (vectorizable_store): Likewise.
18422         (vectorizable_load): Likewise.
18423         (vectorizable_condition): Likewise.
18424         (vectorizable_comparison): Likewise.
18425         (can_vectorize_live_stmts): Likewise.
18426         (vect_analyze_stmt): Likewise.
18427         (vect_transform_stmt): Adjust calls to vectorizable_*.
18428         * tree-vectorizer.c: Include gimple-pretty-print.h.
18429         (dump_stmt_cost): New function.
18431 2018-05-16  Richard Biener  <rguenther@suse.de>
18433         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
18434         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
18435         * tree-ssa-dse.c: Include tree-ssa-loop.h.
18436         (check_name): New callback.
18437         (dse_classify_store): Track cycles via a visited bitmap of PHI
18438         defs and simplify handling of in-loop and across loop dead stores
18439         and properly fail for loop-variant refs.  Handle byte-tracking with
18440         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
18441         limiting the walk.
18443 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
18445         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
18446         (vect_get_mask_type_for_stmt): Likewise.
18447         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
18448         split out from...
18449         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
18450         to determine the statement's vector type and the vector type that
18451         should be used for calculating nunits.  Deal with cases in which
18452         the type has to be deferred.
18453         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
18454         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
18455         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
18456         (vect_determine_vf_for_stmt): New functions, split out from...
18457         (vect_determine_vectorization_factor): ...here.
18458         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
18459         (vect_get_mask_type_for_stmt): New functions, split out from
18460         vect_determine_vectorization_factor.
18462 2018-05-16  Richard Biener  <rguenther@suse.de>
18464         * tree-cfg.c (verify_gimple_assign_ternary): Properly
18465         verify the [VEC_]COND_EXPR embedded comparison.
18467 2018-05-15  Martin Sebor  <msebor@redhat.com>
18469         PR tree-optimization/85753
18470         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
18471         RECORD_TYPE in addition to ARRAY_TYPE.
18473 2018-05-15  Martin Sebor  <msebor@redhat.com>
18475         PR middle-end/85643
18476         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
18478 2018-05-15  Richard Biener  <rguenther@suse.de>
18480         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
18481         add by_clobber_p one.  Change algorithm to collect all defs
18482         representing uses we need to walk and try reducing them to
18483         a single one before failing.
18484         (dse_dom_walker::dse_optimize_stmt): Adjust.
18486 2018-05-13  Mark Wielaard  <mark@klomp.org>
18488         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
18489         (size_of_loc_descr): Likewise.
18490         (output_loc_operands): Likewise.
18491         (output_loc_operands_raw): Likewise.
18492         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
18493         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
18494         (hash_loc_operands): Likewise.
18495         (compare_loc_operands): Likewise.
18497 2018-05-14  Mark Wielaard  <mark@klomp.org>
18499         * dwarf2out.c (count_index_addrs): New function.
18500         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
18502 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18504         PR tree-optimization/83648
18505         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
18506         return value as malloc candidate.
18508 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
18510         PR ipa/85734
18511         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
18512         param as true in call to suggest_attribute.
18514 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
18516         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
18517         -mreadonly-in-sdata.
18519 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18521         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
18522         New pattern.
18523         (aarch64_crypto_aesd_fused): Likewise.
18525 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
18527         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
18528         (movsi_aarch64): Likewise.
18529         (load_pairsi): Likewise.
18530         (load_pairdi): Likewise.
18531         (store_pairsi): Likewise.
18532         (store_pairdi): Likewise.
18533         (load_pairsf): Likewise.
18534         (load_pairdf): Likewise.
18535         (store_pairsf): Likewise.
18536         (store_pairdf): Likewise.
18537         (zero_extend): Likewise.
18538         (trunc): Swap alternatives.
18539         (fcvt_target): Add '?' to prefer w over r.
18541 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
18543         PR target/85756
18544         * config/i386/i386.md: Disallow non-commutative arithmetics in
18545         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
18546         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
18547         in the peephole2 before it.
18549 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
18551         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
18552         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
18553         (ix86_handle_option): Handle -mcldemote.
18554         * config.gcc: New header.
18555         * config/i386/cldemoteintrin.h: New file.
18556         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
18557         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
18558         -mcldemote.
18559         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18560         OPTION_MASK_ISA_CLDEMOTE.
18561         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
18562         (ix86_valid_target_attribute_inner_p): Ditto.
18563         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
18564         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
18565         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
18566         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
18567         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
18568         (cldemote): New.
18569         * config/i386/i386.opt: Add -mcldemote.
18570         * config/i386/x86intrin.h: New header.
18571         * doc/invoke.texi: Add -mcldemote.
18573 2018-05-14  Richard Biener  <rguenther@suse.de>
18575         * doc/match-and-simplify.texi: Adjust :s documentation.
18577 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
18579         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
18580         intended memcpy size.
18581         (REORDER_45): Likewise.
18583 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
18585         * sort.cc: New file.
18586         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
18587         * vec.c (qsort_chk): Use gcc_qsort.
18588         * Makefile.in (OBJS-libcommon): Add sort.o.
18589         (build/sort.o): New target.  Use it...
18590         (BUILD_RTL): ... here, and...
18591         (build/gencfn-macros): ... here, and...
18592         (build/genmatch): ... here.
18594 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
18595             Chung-Ju Wu  <jasonwucj@gmail.com>
18597         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
18598         * config/nds32/nds32-graywolf.md: New file.
18599         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
18600         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
18601         pipeline.
18602         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
18603         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
18604         * config/nds32/nds32.md (pipeline_model): Add graywolf.
18605         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
18606         * config/nds32/pipelines.md: Include n15 settings.
18608 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
18609             Chung-Ju Wu  <jasonwucj@gmail.com>
18611         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
18612         * config/nds32/nds32-n13.md: New file.
18613         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
18614         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
18615         pipeline.
18616         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
18617         * config/nds32/nds32.md (pipeline_model): Add n13.
18618         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
18619         * config/nds32/pipelines.md: Include n13 settings.
18621 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
18622             Chung-Ju Wu  <jasonwucj@gmail.com>
18624         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
18625         * config/nds32/nds32-n10.md: New file.
18626         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
18627         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
18628         pipeline.
18629         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
18630         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
18631         * config/nds32/nds32.md (pipeline_model): Add n10.
18632         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
18633         * config/nds32/pipelines.md: Include n10 settings.
18635 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
18636             Kito Cheng  <kito.cheng@gmail.com>
18637             Chung-Ju Wu  <jasonwucj@gmail.com>
18639         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
18640         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
18641         Add enum values for DSP extension instructions.
18642         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
18643         New constraints.
18644         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
18645         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
18646         New code iterators.
18647         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
18648         * config/nds32/nds32-dspext.md: New file for DSP implementation.
18649         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
18650         * config/nds32/nds32-intrinsic.md: Likewise.
18651         * config/nds32/nds32_intrinsic.h: Likewise.
18652         * config/nds32/nds32-md-auxiliary.c: Likewise.
18653         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
18654         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
18655         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
18656         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
18657         * config/nds32/nds32-protos.h: New declarations for DSP extension.
18658         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
18659         TYPE_DMAC in switch statement.
18660         * config/nds32/nds32.c: New checking and implementation for DSP
18661         extension instructions.
18662         * config/nds32/nds32.h: Likewise.
18663         * config/nds32/nds32.md: Likewise.
18664         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
18665         * config/nds32/predicates.md: Implement new predicates for DSP
18666         extension.
18668 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
18670         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
18671         Reformat alternatives and attributes so it is easier to identify
18672         which constraints/attributes go with which instruction.
18673         (mov<mode>_hardfloat32, FMOVE64): Likewise.
18674         (mov<mode>_softfloat32, FMOVE64): Likewise.
18675         (mov<mode>_hardfloat64, FMOVE64): Likewise.
18676         (mov<mode>_softfloat64, FMOVE64): Likewise.
18678 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18680         * doc/extend.texi (PowerPC Built-in Functions): Rename this
18681         subsection.
18682         (Basic PowerPC Built-in Functions): The new name of the
18683         subsection previously known as "PowerPC Built-in Functions".
18684         (Basic PowerPC Built-in Functions Available on all Configurations):
18685         New subsubsection.
18686         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
18687         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
18688         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
18689         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
18691 2018-05-11  Martin Jambor  <mjambor@suse.cz>
18693         PR ipa/85655
18694         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
18695         single const.
18697 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
18699         PR target/85733
18700         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
18702 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
18704         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
18705         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
18706         (ix86_handle_option): Handle -mwaitpkg.
18707         * config.gcc: New header.
18708         * config/i386/cpuid.h (bit_WAITPKG): New bit.
18709         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
18710         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
18711         function type.
18712         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18713         OPTION_MASK_ISA_WAITPKG.
18714         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
18715         (ix86_option_override_internal): Add PTA_WAITPKG.
18716         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
18717         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
18718         IX86_BUILTIN_TPAUSE.
18719         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
18720         __builtin_ia32_umwait and __builtin_ia32_tpause.
18721         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
18722         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
18723         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
18724         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
18725         UNSPECV_TPAUSE): New.
18726         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
18727         * config/i386/i386.opt: Add -mwaitpkg.
18728         * config/i386/waitpkgintrin.h: New file.
18729         * config/i386/x86intrin.h: New header.
18730         * doc/invoke.texi: Add -mwaitpkg.
18732 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
18734         PR target/85606
18735         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
18736         equivalent.
18737         (cortex-m0): Use armv6s-m isa.
18738         (cortex-m0plus): Likewise.
18739         (cortex-m1): Likewise.
18740         (cortex-m0.small-multiply): Likewise.
18741         (cortex-m0plus.small-multiply): Likewise.
18742         (cortex-m1.small-multiply): Likewise.
18744 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
18745             Jakub Jelinek  <jakub@redhat.com>
18747         PR tree-optimization/85692
18748         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
18749         source permute as well.
18751 2018-05-11  Martin Liska  <mliska@suse.cz>
18753         PR sanitizer/85556
18754         * doc/extend.texi: Document LLVM style format for no_sanitize
18755         attribute.
18757 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
18759         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
18760         mode_supports_vsx_dform_quad to mode_supports_dq_form.
18761         (mode_supports_vsx_dform_quad): Likewise.
18762         (mode_supports_vmx_dform): Move these functions to be next to the
18763         other mode_supports functions.
18764         (mode_supports_dq_form): Likewise.
18765         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
18766         mode_supports_dq_form.
18767         (reg_offset_addressing_ok_p): Likewise.
18768         (offsettable_ok_by_alignment): Likewise.
18769         (rs6000_legitimate_offset_address_p): Likewise.
18770         (legitimate_lo_sum_address_p): Likewise.
18771         (rs6000_legitimize_address): Likewise.
18772         (rs6000_legitimize_reload_address): Likewise.
18773         (rs6000_secondary_reload_inner): Likewise.
18774         (rs6000_preferred_reload_class): Likewise.
18775         (rs6000_output_move_128bit): Likewise.
18777 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
18779         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
18780         Generate SImode target register for null target.
18781         <case IX86_BUILTIN_XGETBV>: Ditto.
18782         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
18783         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
18785 2018-05-10  Carl Love  <cel@us.ibm.com>
18787         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
18788         dcbtt and dcbtstt if operands[2] is 0.
18790 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
18792         PR target/85693
18793         * config/i386/sse.md (usadv64qi): New expander.
18795 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
18797         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
18798         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
18799         -maltivec=be support.
18800         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
18801         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
18802         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
18803         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
18804         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
18805         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
18806         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
18807         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
18808         altivec_vsumsws): Adjust.
18809         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
18810         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
18811         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
18812         support.
18813         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
18814         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
18815         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
18816         (altivec_lve<VI_char>x): Delete expand.
18817         (*altivec_lve<VI_char>x_internal): Rename to...
18818         (altivec_lve<VI_char>x): ... this.
18819         (altivec_lvxl_<mode>): Delete expand.
18820         (*altivec_lvxl_<mode>_internal): Rename to ...
18821         (altivec_lvxl_<mode>): ... this.
18822         (altivec_stvxl_<mode>): Delete expand.
18823         (*altivec_stvxl_<mode>_internal): Rename to ...
18824         (altivec_stvxl_<mode>): ... this.
18825         (altivec_stve<VI_char>x): Delete expand.
18826         (*altivec_stve<VI_char>x_internal): Rename to ...
18827         (altivec_stve<VI_char>x): ... this.
18828         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
18829         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
18830         reduc_plus_scal_<mode>): Adjust.
18831         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
18832         comment.
18833         (rs6000_cpu_cpp_builtins): Adjust.
18834         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
18835         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
18836         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
18837         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
18838         -maltivec=be support.
18839         (rs6000_split_vec_extract_var): Adjust.
18840         (rs6000_split_v4si_init): Adjust.
18841         (swap_selector_for_mode): Delete.
18842         (altivec_expand_lvx_be, altivec_expand_stvx_be,
18843         altivec_expand_stvex_be): Delete.
18844         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
18845         -maltivec=be support.
18846         (rs6000_gimple_fold_builtin): Ditto.
18847         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
18848         Adjust.
18849         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
18850         (TARGET_DIRECT_MOVE_64BIT): Adjust.
18851         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
18852         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
18853         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
18854         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
18855         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
18856         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
18857         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
18858         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
18859         anonymous split): Adjust.
18860         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
18861         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
18863 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
18865         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
18866         when --with-gxx-include-dir is also specified.
18867         * configure: Regenerate.
18869 2018-05-09  Jim Wilson  <jimw@sifive.com>
18871         PR target/84797
18872         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
18873         * config/riscv/t-withmultilib: New.
18874         * config/riscv/withmultilib.h: New.
18875         * doc/install.texi: Document RISC-V --with-multilib-list support.
18877 2018-05-09  Richard Biener  <rguenther@suse.de>
18879         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
18880         vector.
18881         (vect_bb_vectorization_profitable_p): Adjust.  Compute
18882         actual scalar cost using the cost vector and the add_stmt_cost
18883         machinery.
18885 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
18887         PR rtl-optimization/85645
18888         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
18889         in the REG_CFA_REGISTER note for LR, don't leave it empty.
18891 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
18893         PR rtl-optimization/85645
18894         * shrink-wrap.c (spread_components): Return a boolean saying if
18895         anything was changed.
18896         (try_shrink_wrapping_separate): Iterate spread_components until
18897         nothing changes anymore.
18899 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
18901         PR rtl-optimization/85645
18902         * regrename.c (build_def_use): Also kill the chains that include the
18903         destination of a REG_CFA_REGISTER note.
18905 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
18907         PR rtl-optimization/85645
18908         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
18909         insn that has a REG_CFA_REGISTER note.
18911 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
18913         * cfgexpand.c (expand_clobber): New function.
18914         (expand_gimple_stmt_1): Use it.
18915         * tree-vect-stmts.c (vect_clobber_variable): New function,
18916         split out from...
18917         (vectorizable_simd_clone_call): ...here.
18918         (vectorizable_store): Emit a clobber either side of an
18919         IFN_STORE_LANES sequence.
18920         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
18922 2018-05-09  Tom de Vries  <tom@codesourcery.com>
18924         PR target/85626
18925         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
18926         (define_insn "trap_if_false"): Add exit after trap.
18928 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
18930         PR rtl-optimization/85638
18931         * bb-reorder.c: Include common/common-target.h.
18932         (create_forwarder_block): New function extracted from...
18933         (fix_up_crossing_landing_pad): ...here.  Rename into...
18934         (dw2_fix_up_crossing_landing_pad): ...this.
18935         (sjlj_fix_up_crossing_landing_pad): New function.
18936         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
18937         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
18938         from both partitions and exit the loop after one iteration.
18940 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18942         Revert:
18943         * doc/extend.texi (PowerPC Built-in Functions): Rename this
18944         subsection.
18945         (Basic PowerPC Built-in Functions): The new name of the
18946         subsection previously known as "PowerPC Built-in Functions".
18947         (Basic PowerPC Built-in Functions Available on all Configurations):
18948         New subsubsection.
18949         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
18950         subsubsection.
18951         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
18952         subsubsection.
18953         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
18954         subsubsection.
18955         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
18956         subsubsection.
18958 2018-05-08  Jim Wilson  <jimw@sifive.com>
18960         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
18961         (LD_EMUL_SUFFIX): New.
18962         (LINK_SPEC): Use it.
18964 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18966         * doc/extend.texi (PowerPC Built-in Functions): Rename this
18967         subsection.
18968         (Basic PowerPC Built-in Functions): The new name of the
18969         subsection previously known as "PowerPC Built-in Functions".
18970         (Basic PowerPC Built-in Functions Available on all Configurations):
18971         New subsubsection.
18972         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
18973         subsubsection.
18974         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
18975         subsubsection.
18976         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
18977         subsubsection.
18978         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
18979         subsubsection.
18981 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
18983         PR target/85683
18984         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
18985         after cmpelim optimization.
18987 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
18989         * config.gcc: Support "goldmont".
18990         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
18991         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18992         PROCESSOR_GOLDMONT.
18993         * config/i386/i386.c (m_GOLDMONT): Define.
18994         (processor_target_table): Add "goldmont".
18995         (PTA_GOLDMONT): Define.
18996         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
18997         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
18998         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
18999         (fold_builtin_cpu): Add "goldmont".
19000         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
19001         (ix86_option_override_internal): Add "goldmont".
19002         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
19003         (processor_type): Add PROCESSOR_GOLDMONT.
19004         * config/i386/i386.md: Add CPU "glm".
19005         * config/i386/glm.md: New file.
19006         * config/i386/x86-tune.def: Add m_GOLDMONT.
19007         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
19009 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
19011         PR target/85572
19012         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
19013         E_V4DImode.
19014         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
19015         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
19016         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
19018         PR target/85317
19019         * config/i386/i386.c (ix86_fold_builtin): Handle
19020         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
19022         PR target/85480
19023         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
19024         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
19026 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
19028         PR target/85658
19029         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
19030         (check_arch): Likewise.
19031         (check_fpu): Return the result rather than printing it.
19032         (end arch): Fix operator precedence.
19033         (end cpu): Likewise.
19034         (END): Print the result from check_fpu.
19036 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
19037             Alan Hayward  <alan.hayward@arm.com>
19038             David Sherwood  <david.sherwood@arm.com>
19040         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
19041         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
19042         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
19043         (*fcmuo<mode>_and): New patterns.
19045 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
19047         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
19048         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
19049         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
19050         (cmp_op, sve_imm_con): New code attributes.
19051         (SVE_COND_INT_CMP, imm_con): Delete.
19052         (cmp_op): Remove above unspecs from int attribute.
19053         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
19054         to...
19055         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
19056         comparison-specific unspecs.
19057         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
19058         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
19059         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
19060         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
19061         (*vec_fcm<cmp_op><mode>): Rename to...
19062         (*fcm<cmp_op><mode>): ...this and adjust likewise.
19063         (*vec_fcmuo<mode>): Rename to...
19064         (*fcmuo<mode>): ...this and adjust likewise.
19065         (*pred_fcm<cmp_op><mode>): New pattern.
19066         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
19067         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
19068         functions.
19069         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
19070         and UNORDERED.
19071         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
19072         (aarch64_emit_sve_predicated_cond): New function.
19073         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
19074         (aarch64_emit_unspec_cond_or): Replace with...
19075         (aarch64_emit_sve_or_conds): ...this new function.  Use
19076         aarch64_emit_sve_ptrue_op for the individual comparisons and
19077         aarch64_emit_binop to OR them together.
19078         (aarch64_emit_inverted_unspec_cond): Replace with...
19079         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
19080         aarch64_emit_sve_ptrue_op for the comparison and
19081         aarch64_emit_unop to invert the result.
19082         (aarch64_expand_sve_vec_cmp_float): Update after the above
19083         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
19085 2018-05-07  Nathan Sidwell  <nathan@acm.org>
19087         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
19088         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
19089         (Backwards Compatibility): Likewise.
19091 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19093         PR bootstrap/85681
19094         Revert:
19095         2018-05-07  Luis Machado  <luis.machado@linaro.org>
19097         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19098         <prefetch_dynamic_strides>: New const bool field.
19099         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19100         prefetch_dynamic_strides.
19101         (exynosm1_prefetch_tune): Likewise.
19102         (thunderxt88_prefetch_tune): Likewise.
19103         (thunderx_prefetch_tune): Likewise.
19104         (thunderx2t99_prefetch_tune): Likewise.
19105         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
19106         to false.
19107         (aarch64_override_options_internal): Update to set
19108         PARAM_PREFETCH_DYNAMIC_STRIDES.
19109         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19110         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
19111         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
19112         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19113         prefetch-dynamic-strides setting.
19115         2018-05-07  Luis Machado  <luis.machado@linaro.org>
19117         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19118         <minimum_stride>: New const int field.
19119         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19120         minimum_stride field.
19121         (exynosm1_prefetch_tune): Likewise.
19122         (thunderxt88_prefetch_tune): Likewise.
19123         (thunderx_prefetch_tune): Likewise.
19124         (thunderx2t99_prefetch_tune): Likewise.
19125         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
19126         (aarch64_override_options_internal): Update to set
19127         PARAM_PREFETCH_MINIMUM_STRIDE.
19128         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
19129         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
19130         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
19131         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
19132         stride is constant and is below the minimum stride threshold.
19134 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19136         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
19137         to 512.
19139 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19141         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19142         <prefetch_dynamic_strides>: New const bool field.
19143         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19144         prefetch_dynamic_strides.
19145         (exynosm1_prefetch_tune): Likewise.
19146         (thunderxt88_prefetch_tune): Likewise.
19147         (thunderx_prefetch_tune): Likewise.
19148         (thunderx2t99_prefetch_tune): Likewise.
19149         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
19150         to false.
19151         (aarch64_override_options_internal): Update to set
19152         PARAM_PREFETCH_DYNAMIC_STRIDES.
19153         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19154         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
19155         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
19156         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19157         prefetch-dynamic-strides setting.
19159 2018-05-07  Luis Machado  <luis.machado@linaro.org>
19161         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19162         <minimum_stride>: New const int field.
19163         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19164         minimum_stride field.
19165         (exynosm1_prefetch_tune): Likewise.
19166         (thunderxt88_prefetch_tune): Likewise.
19167         (thunderx_prefetch_tune): Likewise.
19168         (thunderx2t99_prefetch_tune): Likewise.
19169         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
19170         (aarch64_override_options_internal): Update to set
19171         PARAM_PREFETCH_MINIMUM_STRIDE.
19172         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
19173         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
19174         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
19175         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
19176         stride is constant and is below the minimum stride threshold.
19178 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
19180         PR c++/85659
19181         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
19182         the type is addressable.  Don't force op into register if it has
19183         BLKmode.
19185 2018-05-05  Roland McGrath  <mcgrathr@google.com>
19187         PR other/77609
19188         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
19189         any section for which we don't know a specific type it should have,
19190         regardless of name.  Previously this was done only for the exact
19191         names ".init_array", ".fini_array", and ".preinit_array".
19192         (default_elf_asm_named_section): Add comment about
19193         relationship with default_section_type_flags and SECTION_NOTYPE.
19194         (get_section): Don't consider it a type conflict if one side has
19195         SECTION_NOTYPE and the other doesn't, as long as neither has the
19196         SECTION_BSS et al used in the default_section_type_flags logic.
19198 2018-05-05  Tom de Vries  <tom@codesourcery.com>
19200         PR target/85653
19201         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
19202         (workaround_barsyncs): New function.
19203         (nvptx_reorg): Use workaround_barsyncs.
19204         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
19205         (define_expand "nvptx_membar_cta"): New define_expand.
19206         (define_insn "*nvptx_membar_cta"): New insn.
19208 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
19210         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
19211         To improve optimization opportunities.
19212         * builtin-types.def: The new needed builtin types for the above.
19214 2018-05-04  Richard Biener  <rguenther@suse.de>
19216         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
19217         * gimple-ssa-store-merging.c
19218         (imm_store_chain_info::output_merged_store): Remove redundant create,
19219         release split_store vector contents on failure.
19220         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
19221         scalar stmt vector on cache hit.
19223 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
19225         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
19226         Xilinx FP support.
19227         * config.gcc (powerpc-xilinx-eabi*): Remove.
19228         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
19229         support.
19230         (fusion_addis_mem_combo_load): Ditto.
19231         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
19232         FP support.
19233         (rs6000_cpu_cpp_builtins): Ditto.
19234         * config/rs6000/rs6000-linux.c
19235         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
19236         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
19237         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
19238         support.
19239         (rs6000_setup_reg_addr_masks): Ditto.
19240         (rs6000_init_hard_regno_mode_ok): Ditto.
19241         (rs6000_option_override_internal): Ditto.
19242         (legitimate_lo_sum_address_p): Ditto.
19243         (rs6000_legitimize_address): Ditto.
19244         (rs6000_legitimize_reload_address): Ditto.
19245         (rs6000_legitimate_address_p): Ditto.
19246         (abi_v4_pass_in_fpr): Ditto.
19247         (setup_incoming_varargs): Ditto.
19248         (rs6000_gimplify_va_arg): Ditto.
19249         (rs6000_split_multireg_move): Ditto.
19250         (rs6000_savres_strategy): Ditto.
19251         (rs6000_emit_prologue_components): Ditto.
19252         (rs6000_emit_epilogue_components): Ditto.
19253         (rs6000_emit_prologue): Ditto.
19254         (rs6000_emit_epilogue): Ditto.
19255         (rs6000_elf_file_end): Ditto.
19256         (rs6000_function_value): Ditto.
19257         (rs6000_libcall_value): Ditto.
19258         * config/rs6000/rs6000.h: Ditto.
19259         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
19260         (TARGET_MINMAX): ... this.  New.
19261         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
19262         * config/rs6000/rs6000.md: Remove Xilinx FP support.
19263         (*movsi_internal1_single): Delete.
19264         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
19265         mfpu=, mxilinx-fpu): Delete.
19266         * config/rs6000/singlefp.h: Delete.
19267         * config/rs6000/sysv4.h: Remove Xilinx FP support.
19268         * config/rs6000/t-rs6000: Ditto.
19269         * config/rs6000/t-xilinx: Delete.
19270         * config/rs6000/titan.md: Adjust for fp_type removal.
19271         * config/rs6000/vsx.md: Remove Xilinx FP support.
19272         (VStype_simple): Delete.
19273         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
19274         * config/rs6000/xfpu.h: Delete.
19275         * config/rs6000/xfpu.md: Delete.
19276         * config/rs6000/xilinx.h: Delete.
19277         * config/rs6000/xilinx.opt: Delete.
19278         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
19279         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
19281 2018-05-04  Tom de Vries  <tom@codesourcery.com>
19283         PR libgomp/85639
19284         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
19285         if ignore == 0.
19287 2018-05-04  Richard Biener  <rguenther@suse.de>
19289         PR middle-end/85627
19290         * tree-complex.c (update_complex_assignment): We are always in SSA form.
19291         (expand_complex_div_wide): Likewise.
19292         (expand_complex_operations_1): Likewise.
19293         (expand_complex_libcall): Preserve EH info of the original stmt.
19294         (tree_lower_complex): Handle removed blocks.
19295         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
19296         on complex multiplication and division libcall builtins.
19298 2018-05-04  Richard Biener  <rguenther@suse.de>
19300         PR middle-end/85574
19301         * fold-const.c (negate_expr_p): Restrict negation of operand
19302         zero of a division to when we know that can happen without
19303         overflow.
19304         (fold_negate_expr_1): Likewise.
19306 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
19308         PR libstdc++/85466
19309         * real.h (real_nextafter): Declare.
19310         * real.c (real_nextafter): New function.
19311         * fold-const-call.c (fold_const_nextafter): New function.
19312         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
19313         CASE_CFN_NEXTTOWARD.
19314         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
19315         even when arg1_mode is different from arg0_mode.
19317 2018-05-03  Nathan Sidwell  <nathan@acm.org>
19319         * doc/extend.texi (Deprecated Features): Remove
19320         -ffriend-injection.
19321         (Backwards Compatibility): Likewise.
19322         * doc/invoke.texi (C++ Language Options): Likewise.
19323         (C++ Dialect Options): Likewise.
19325 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
19327         PR target/85530
19328         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
19329         _mm512_mask_mullox_epi64): New intrinsics.
19331 2018-05-03  Tom de Vries  <tom@codesourcery.com>
19333         PR testsuite/85106
19334         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
19335         dump files): Add offload-tree.
19337 2018-05-03  Richard Biener  <rguenther@suse.de>
19339         PR tree-optimization/85615
19340         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
19341         to loops not nested in BBs loop father to avoid creating multi-entry
19342         loops.
19344 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19346         PR tree-optimization/70291
19347         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
19348         arguments.  Change return type to tree.  Emit libcall as a new
19349         statement rather than replacing existing one when inplace_p is true.
19350         (expand_complex_multiplication_components): New function.
19351         (expand_complex_multiplication): Expand floating-point complex
19352         multiplication using the above.
19353         (expand_complex_division): Rename inner_type parameter to type.
19354         Update expand_complex_libcall call-site.
19355         (expand_complex_operations_1): Update expand_complex_multiplication
19356         and expand_complex_division call-sites.
19358 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
19360         PR target/85582
19361         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
19362         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
19363         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
19364         the highest significant bit of the shift count mask is clear.  In
19365         check whether and[sq]i3 is needed verify that all significant bits
19366         of the shift count other than the highest are set.
19368 2018-05-02  Tom de Vries  <tom@codesourcery.com>
19370         PR libgomp/82428
19371         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
19372         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
19373         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
19374         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
19375         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
19376         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
19377         __builtin_goacc_parlevel_size.
19379 2018-05-02  Richard Biener  <rguenther@suse.de>
19381         PR tree-optimization/85597
19382         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
19383         do not use split vect_get_vec_defs call but call vect_get_slp_defs
19384         directly.
19386 2018-05-02  Tom de Vries  <tom@codesourcery.com>
19388         PR testsuite/85106
19389         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
19390         dump files): Add ltrans-tree.
19392 2018-05-02  Tom de Vries  <tom@codesourcery.com>
19394         PR testsuite/85106
19395         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
19396         dump files): Add wpa-ipa.
19398 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
19400         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
19401         powerpc*-*-linux*paired* target.
19402         * config/rs6000/750cl.h: Delete.
19403         * config/rs6000/paired.h: Delete.
19404         * config/rs6000/paired.md: Delete.
19405         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
19406         float support.
19407         * config/rs6000/rs6000-builtin.def: Remove paired float support.
19408         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
19409         comment.  Remove paired float support.
19410         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
19411         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
19412         VECTOR_PAIRED.
19413         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
19414         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
19415         declarations.
19416         * config/rs6000/rs6000.c: Remove paired float support.
19417         (paired_expand_vector_init, paired_expand_vector_move,
19418         paired_emit_vector_compare, paired_emit_vector_cond_expr,
19419         (paired_expand_lv_builtin, paired_expand_stv_builtin,
19420         paired_expand_builtin, paired_expand_predicate_builtin,
19421         paired_init_builtins): Delete.
19422         * config/rs6000/rs6000.h: Remove paired float support.
19423         * config/rs6000/rs6000.md: Remove paired float support.
19424         (move_from_CR_ov_bit): Delete.
19425         * config/rs6000/rs6000.opt (mpaired): Delete.
19426         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
19427         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
19429 2018-05-02  Richard Biener  <rguenther@suse.de>
19431         PR middle-end/85567
19432         * gimplify.c (gimplify_save_expr): When in SSA form allow
19433         SAVE_EXPRs to compute to SSA vars.
19435 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
19437         PR target/85582
19438         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
19439         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
19440         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
19441         clobber operands[2], instead use a new pseudo.  Formatting fixes.
19443 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
19445         PR tree-optimization/85586
19446         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
19447         exit early for statements in the same group if the accesses are
19448         not strided.
19450 2018-05-02  Tom de Vries  <tom@codesourcery.com>
19452         PR lto/85451
19453         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
19454         error message.
19456 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
19458         PR tree-optimization/85143
19459         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
19461 2018-05-01  Tom de Vries  <tom@codesourcery.com>
19463         PR lto/85451
19464         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
19465         not found" error message.
19467 2018-05-01  Tom de Vries  <tom@codesourcery.com>
19469         PR other/83786
19470         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
19471         * vec.c (test_ordered_remove_if): New function.
19472         (vec_c_tests): Call test_ordered_remove_if.
19473         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
19474         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
19475         * tree-vect-patterns.c (vect_pattern_recog_1): Use
19476         VEC_ORDERED_REMOVE_IF.
19478 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19480         PR tree-optimization/82665
19481         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
19482         pointer subtraction where arguments come from a memchr call.
19484 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
19486         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
19487         --push-state --as-needed and --pop-state instead of --as-needed and
19488         --no-as-needed if ld supports it.
19489         * configure: Regenerated.
19491         PR web/85578
19492         * doc/install.texi2html: Replace _002d with - and _002a with * in
19493         generated html files using sed.
19495 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
19497         PR c++/85523
19498         * gcc-rich-location.c (blank_line_before_p): New function.
19499         (use_new_line): New function.
19500         (gcc_rich_location::add_fixit_insert_formatted): New function.
19501         * gcc-rich-location.h
19502         (gcc_rich_location::add_fixit_insert_formatted): New function.
19504 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
19506         * selftest.c (assert_streq): Rename "expected" and "actual" to
19507         "val1" and "val2".  Extend NULL-handling to cover both inputs
19508         symmetrically, while still requiring both to be non-NULL for a pass.
19509         * selftest.h (assert_streq): Rename "expected" and "actual" to
19510         "val1" and "val2".
19511         (ASSERT_EQ): Likewise.
19512         (ASSERT_EQ_AT): Likewise.
19513         (ASSERT_KNOWN_EQ): Likewise.
19514         (ASSERT_KNOWN_EQ_AT): Likewise.
19515         (ASSERT_NE): Likewise.
19516         (ASSERT_MAYBE_NE): Likewise.
19517         (ASSERT_MAYBE_NE_AT): Likewise.
19518         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
19519         the assertion to pass.
19520         (ASSERT_STREQ_AT): Likewise.
19522 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
19524         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
19525         interaction with -pie.
19527 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
19529         * selftest.h: Fix alphabetization of per-source-file selftest
19530         declarations.
19532 2018-04-30  Jason Merrill  <jason@redhat.com>
19534         PR c++/61982 - dead stores to destroyed objects.
19535         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
19536         of clobber.
19538 2018-04-30  Jason Merrill  <jason@redhat.com>
19540         * tree.c (build_clobber): New.
19541         * tree.h: Declare it.
19542         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
19544 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
19546         * diagnostic-show-locus.c (layout::layout): Update for
19547         location_get_source_line returning a char_span.
19548         (struct char_span): Move to input.h.
19549         (struct correction): Update for fields in char_span becoming
19550         private.
19551         (struct source_line): Update for location_get_source_line
19552         returning a char_span.
19553         (layout::print_line): Likewise.
19554         * edit-context.c (edited_file::print_content): Likewise.
19555         (edited_file::print_diff_hunk): Likewise.
19556         (edited_file::print_run_of_changed_lines): Likewise.
19557         (edited_file::get_num_lines): Likewise.
19558         (edited_line::edited_line): Likewise.
19559         * final.c (asm_show_source): Likewise.
19560         * input.c (location_get_source_line): Convert return type
19561         from const char * to char_span, losing the final "line_len"
19562         param.
19563         (dump_location_info): Update for the above.
19564         (get_substring_ranges_for_loc): Likewise.  Use a char_span
19565         when handling the literal within the line.
19566         (test_reading_source_line): Update for location_get_source_line
19567         returning a char_span.
19568         * input.h (class char_span): Move here from
19569         diagnostic-show-locus.c, converting from a struct to a class.
19570         Make data members private.
19571         (char_span::operator bool): New.
19572         (char_span::length): New.
19573         (char_span::get_buffer): New.
19574         (char_span::operator[]): New.
19575         (char_span::subspan): Make const.
19576         (char_span::xstrdup): New.
19577         (location_get_source_line): Convert return type from const char *
19578         to char_span, losing the final "line_size" param.
19580 2018-04-30  Jan Hubicka  <jh@suse.cz>
19582         * lto-wrapper.c (ltrans_priorities): New static var.
19583         (cmp_priority): New.
19584         (run_gcc): Read priorities and if doing parallel build order
19585         the Makefile by them.
19587 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
19589         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
19591 2018-04-30  Richard Biener  <rguenther@suse.de>
19593         * tree-cfg.c (verify_address): Remove base argument, add
19594         flag whether to check TREE_ADDRESSABLE and do that.
19595         (verify_expr): Remove.
19596         (verify_types_in_gimple_reference): Add pieces from verify_expr.
19597         (verify_gimple_assign_single): Likewise.
19598         (verify_gimple_switch): Likewise.
19599         (verify_expr_location_1): Dereference tp once.  Add (disabled)
19600         piece from verify_expr.
19601         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
19603 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
19605         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
19607 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
19609         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
19610         (small_data_pattern): Likewise.
19611         (arc_rewrite_small_data): Likewise.
19612         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
19613         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
19614         (get_symbol_alignment): New function.
19615         (legitimate_small_data_address_p): Likewise.
19616         (legitimate_scaled_address): Update, call
19617         legitimate_small_data_address_p.
19618         (output_sdata): New static variable.
19619         (arc_print_operand): Update how we handle small data operands.
19620         (arc_print_operand_address): Likewise.
19621         (arc_legitimate_address_p): Update, use
19622         legitimate_small_data_address_p.
19623         (arc_rewrite_small_data_p): Remove.
19624         (arc_rewrite_small_data_1): Likewise.
19625         (arc_rewrite_small_data): Likewise.
19626         (small_data_pattern): Likewise.
19627         (compact_sda_memory_operand): Update to use
19628         legitimate_small_data_address_p and get_symbol_alignment.
19629         (prepare_move_operands): Don't rewite sdata pattern.
19630         (prepare_extend_operands): Remove.
19631         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
19632         pattern.
19633         (zero_extendqisi2): Likewise.
19634         (zero_extendhisi2): Likewise.
19635         (extendqihi2): Likewise.
19636         (extendqisi2): Likewise.
19637         (extendhisi2): Likewise.
19638         (addsi3): Likewise.
19639         (subsi3): Likewise.
19640         (andsi3): Likewise.
19641         * config/arc/constraints.md (Usd): Change it to memory constraint.
19643 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
19645         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
19646         as source of std instructions.
19647         * config/arc/arc.md (movsi_insn): Update pattern predicate to
19648         allow 6-bit constants as source for store instructions.
19649         (movdi_insn): Update instruction pattern to allow 6-bit constants
19650         as source for store instructions.
19652 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
19654         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
19656 2018-04-30  Nathan Sidwell  <nathan@acm.org>
19657             Sandra Loosemore <sandra@codesourcery.com>
19659         * dumpfile.c (dump_open): Allow '-' for stdout.
19660         * doc/invoke.texi (Developer Options): Document dump filename
19661         determination early.  Document stdin/stdout selection.
19663 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
19665         Microblaze Target: PIC data text relative
19667         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
19668         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
19669         Add declaration.
19670         * config/microblaze/microblaze.h (microblaze_constant_address_p):
19671         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
19672         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
19673         New addressing mode for data-text relative position indepenedent code.
19674         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
19675         'ADDRESS_SYMBOLIC_TXT_REL'.
19676         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
19677         (microblaze_legitimate_pic_operand): Exclude function calls from
19678         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
19679         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
19680         addresses cases.
19681         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
19682         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
19683         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
19684         for 'address + offset'.
19685         (microblaze_expand_prologue): Add new function prologue call for
19686         'r20' assignation.
19687         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
19688         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
19689         table in case of TARGET_PIC_DATA_TEXT_REL.
19690         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
19691         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
19692         Add new macros 'UNSPEC_TEXT',
19693         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
19694         + exclude function calls from 'UNSPEC_PLT' in case of data text
19695         relative mode.
19696         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
19697         new target hook for generating address diff vector tables in case of
19698         flag_pic.
19699         * doc/tm.texi : Regenerate.
19700         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
19701         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
19702         of addr diff vector generation.
19703         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
19704         target hook definition.
19705         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
19706         Add default function for generate_pic_addr_diff_vec -> flag_pic.
19707         * doc/invoke.texi (Add new pic option): Add new microblaze pic
19708         option for data text relative.
19710 2018-04-30  Richard Biener  <rguenther@suse.de>
19712         * tree-chrec.h (evolution_function_is_constant_p): Remove
19713         redundant check.
19714         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
19716 2018-04-30  Richard Biener  <rguenther@suse.de>
19718         PR bootstrap/85571
19719         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
19721 2018-04-30  Richard Biener  <rguenther@suse.de>
19723         PR tree-optimization/28364
19724         PR tree-optimization/85275
19725         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
19726         copying first exit test.
19728 2018-04-28  Mark Wielaard  <mark@klomp.org>
19730         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
19731         dwarf_version >= 5.
19732         (dwarf_AT): Handle DW_AT_addr_base.
19733         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
19735 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
19737         PR target/84431
19738         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
19739         (*ashl<dwi>3_doubleword_mask_1): Ditto.
19740         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
19741         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
19743 2018-04-28  Richard Biener  <rguenther@suse.de>
19745         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
19746         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
19747         to reflect use.  Only add interesting stmts.
19749 2018-04-27  Martin Jambor  <mjambor@suse.cz>
19751         PR ipa/85549
19752         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
19753         the jump function allows for passing through aggregate values.
19755 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
19757         * input.h (in_system_header_at): Convert from macro to inline
19758         function.
19759         (from_macro_expansion_at): Likewise.
19760         (from_macro_definition_at): Likewise.
19762 2018-04-27  Jeff Law  <law@redhat.com>
19764         * config.gcc: Mark tile* targets as deprecated/obsolete.
19766 2018-04-27  Richard Biener  <rguenther@suse.de>
19768         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
19769         fix for ILP32.
19771 2018-04-27  Richard Biener  <rguenther@suse.de>
19773         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
19775 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
19777         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
19778         with Yd constraint. Set "preferred_for_speed" attribute from
19779         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
19780         with Yd constraint.
19781         (*movdi_internal): Ditto.
19782         (movti_interunit splitters): Remove
19783         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
19784         (movdi_interunit splitters): Ditto.
19785         * config/i386/constraints.md (Ye): Remove.
19786         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
19788 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19790         PR target/85512
19791         * config/aarch64/constraints.md (Usg): Limit to 31.
19792         (Usj): Limit to 63.
19794 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
19796         PR tree-optimization/85529
19797         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
19798         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
19799         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
19800         zero extension or masking of the MSB bit.
19801         (optimize_range_tests): Add FIRST_BB argument, pass it through
19802         to optimize_range_tests_var_bound.
19803         (maybe_optimize_range_tests, reassociate_bb): Adjust
19804         optimize_range_tests callers.
19806 2018-04-26  Richard Biener  <rguenther@suse.de>
19807             Jakub Jelinek  <jakub@redhat.com>
19809         * cgraph.h (symbol_table): Just declare debug method here.
19810         * symtab.c (symbol_table::debug): Define.
19812 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
19814         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
19816 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
19818         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
19819         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
19820         (*movdi_internal): Substitute Yi and Yj constraint with x
19821         and Ym and Yn constraint with y constraint.  Update "isa"
19822         attribute and set "preferred_for_speed" attribute from
19823         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
19824         (*movsi_internal): Ditto.
19825         (*movdf_internal): Ditto.
19826         (*movsf_internal): Ditto.
19827         (*zero_extendsidi2): Ditto.
19828         * config/i386/sse.md (vec_set<mode>_0): Ditto.
19829         (sse2_loadld): Ditto.
19830         (*vec_extract<ssevecmodelower>_0): Ditto.
19831         (*vec_extractv4si_0_zext_sse4): Ditto.
19832         (vec_concatv2di): Ditto.
19833         (*vec_dup<mode>): Ditto.
19834         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
19835         * config/i386/constraints.md (Yi): Remove.
19836         (Yj): Remove.
19837         (Ym): Remove.
19838         (Yn): Remove.
19840 2018-04-26  Nathan Sidwell  <nathan@acm.org>
19842         * dumpfile.c (dump_open): New.
19843         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
19844         (dump_finish): Detect stdio/stderr by value not name.
19846 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
19848         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
19850 2018-04-26  Tom de Vries  <tom@codesourcery.com>
19852         PR target/84952
19853         * config/nvptx/nvptx.c (verify_neutering_jumps)
19854         (verify_neutering_labels): New function
19855         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
19857 2018-04-26  Tom de Vries  <tom@codesourcery.com>
19859         PR target/84025
19860         * config/nvptx/nvptx.c (needs_neutering_p): New function.
19861         (nvptx_single): Use needs_neutering_p to skip over insns that do not
19862         need neutering.
19864 2018-04-26  Richard Biener <rguenther@suse.de>
19865             Tom de Vries  <tom@codesourcery.com>
19867         PR lto/85422
19868         * lto-streamer-out.c (output_function): Fixup loops if required to match
19869         discovery done in the reader.
19871 2018-04-26  Richard Biener  <rguenther@suse.de>
19873         PR tree-optimization/85116
19874         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
19875         have a loop exit from the single latch predecessor.  Remove
19876         case of header with just condition.
19877         (ch_base::copy_headers): Exclude infinite loops from any
19878         processing.
19879         (pass_ch::execute): Record exits.
19881 2018-04-26  Richard Biener  <rguenther@suse.de>
19883         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
19884         prologue cost vector and pass it to vect_get_load_cost.
19885         (vect_get_peeling_costs_all_drs): Likewise.
19886         (vect_peeling_hash_get_lowest_cost): Likewise.
19887         (vect_enhance_data_refs_alignment): Likewise.
19889 2018-04-26  Richard Biener  <rguenther@suse.de>
19891         PR middle-end/85450
19892         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
19893         checking of integer<->pointer conversions.
19894         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
19895         sign-/zero-extending pointer types.
19896         (expand_omp_for_static_chunk): Likewise.
19898 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
19899             Jean Lee  <xiaoyur347@gmail.com>
19901         * config/mips/mips.c (mips_asan_shadow_offset): New function.
19902         (TARGET_ASAN_SHADOW_OFFSET): Define.
19903         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
19904         true for -fsanitize=address.
19906 2018-04-25  Mark Wielaard  <mark@klomp.org>
19908         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
19909         shorter ones.
19911 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
19913         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
19914         than "alu", remove explicit "memory" and "imm_disp" attributes.
19915         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
19917         PR middle-end/85414
19918         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
19919         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
19920         gen_lowpart_no_emit.
19922 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
19924         PR target/85473
19925         * config/i386/i386.c (ix86_expand_builtin): Change memory
19926         operand to XI, extend p0 to Pmode.
19927         * config/i386/i386.md: Change unspec volatile and operand
19928         1 mode to XI, change operand 0 mode to P.
19930 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
19932         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
19933         GET_MODE_MASK before any checking.
19934         (nds32_can_use_bset_p): Likewise.
19935         (nds32_can_use_btgl_p): Likewise.
19937 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
19939         * config/nds32/nds32-doubleword.md: New define_split pattern for
19940         illegal register number.
19942 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
19944         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
19946 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
19948         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
19950 2018-04-25  Richard Biener  <rguenther@suse.de>
19952         * lto-streamer.h (LTO_major_version): Bump to 8.
19954 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
19956         * BASE-VER: Set to 9.0.0.
19958 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
19960         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
19961         in __abskf2 and __powikf2.
19963 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19965         PR target/85512
19966         * config/aarch64/constraints.md (Usg, Usj): New constraints.
19967         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
19968         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
19969         Use the above on operand 2.  Reindent.
19970         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
19972 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
19974         PR target/85485
19975         * common/config/i386/i386-common.c (ix86_handle_option): Don't
19976         handle OPT_mcet.
19977         * config/i386/i386.opt (mcet): Removed.
19978         * doc/install.texi: Remove -mcet documentation.
19979         * doc/invoke.texi: Likewise.
19981 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
19983         PR target/85485
19984         * doc/install.texi: Remove -mcet from bootstrap-cet.
19986 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
19988         PR target/85511
19989         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
19990         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
19991         if TARGET_64BIT.
19993         PR target/85503
19994         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
19995         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
19996         containing a CONST_VECTOR.
19998 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
20000         * doc/install.texi: Update newlib dependency for nvptx.
20002 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20004         PR target/85508
20005         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
20006         instead of INTVAL when shifting x left.
20008 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
20010         PR tree-optimization/85478
20011         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
20012         vect_grouped_store_supported for single element vectors.
20014 2018-04-24  Richard Biener  <rguenther@suse.de>
20016         PR target/85491
20017         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
20018         load cost increase to the case of non-constant step.
20020 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
20022         PR target/84828
20023         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
20024         destination if any_malformed_asm.
20026 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
20028         PR middle-end/85496
20029         * expr.c (store_field): In the bitfield case, if the value comes from
20030         a function call and is returned in registers by means of a PARALLEL,
20031         do not change the mode of the temporary unless BLKmode and VOIDmode.
20033 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
20035         PR rtl-optimization/85423
20036         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
20037         dependencies to debug insns when the previous insn is non-debug.
20039 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
20041         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
20042         enums into a single definition.
20043         (fls): Fix predicates and printing.
20044         (seti): Likewise.
20046 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
20048         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
20049         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
20050         and short u6 immediate.
20051         (check_if_valid_sleep_operand): Remove.
20052         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
20054 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20056         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
20057         flag_always_save_lp condition.
20058         * config/nds32/nds32.opt (malways-save-lp): New option.
20060 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
20062         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
20063         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
20064         * config/nds32/nds32.h
20065         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
20066         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
20068 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
20070         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
20071         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
20073 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
20074             Chung-Ju Wu  <jasonwucj@gmail.com>
20076         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
20077         Declare.
20078         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
20079         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
20081 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20083         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
20085 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20087         * config/nds32/nds32-protos.h (nds32_data_alignment,
20088         nds32_local_alignment): Declare.
20089         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
20090         nds32_local_alignment): New functions.
20091         (TARGET_CONSTANT_ALIGNMENT): Define.
20092         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
20094 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20096         * config/nds32/nds32.c
20097         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
20098         (TARGET_MODES_TIEABLE_P): Likewise.
20100 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
20102         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
20103         level Ofast and Og.
20105 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
20106             Chung-Ju Wu  <jasonwucj@gmail.com>
20108         * config/nds32/constants.md (unspec_volatile_element): Add enum values
20109         for unaligned access.
20110         * config/nds32/nds32-intrinsic.c: Implementation of expanding
20111         unaligned access.
20112         * config/nds32/nds32-intrinsic.md: Likewise.
20113         * config/nds32/nds32_intrinsic.h: Likewise.
20114         * config/nds32/nds32.h (nds32_builtins): Likewise.
20115         * config/nds32/nds32.opt (munaligned-access): New option.
20116         * config/nds32/nds32.c (nds32_asm_file_start): Display
20117         flag_unaligned_access status.
20119 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
20121         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
20122         -mno-relax is present.
20123         * config/riscv/linux.h (LINK_SPEC): Ditto.
20125 2018-04-20  Martin Sebor  <msebor@redhat.com>
20127         PR c/85365
20128         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
20129         for null pointers.
20130         (gimple_fold_builtin_stxcpy_chk): Same.
20131         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
20133 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
20135         PR target/85456
20136         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
20137         __powikf2 when long double is IEEE 128-bit.
20139 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
20141         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
20142         step to make sure stack always aligned.
20144 2018-04-20  Carl Love  <cel@us.ibm.com>
20146         PR target/83402
20147         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
20148         size check for arg0.
20150 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
20151             Tom de Vries  <tom@codesourcery.com>
20153         PR target/85445
20154         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
20155         Emit insns for calls too.
20156         (nvptx_find_par): Always look for worker-level predecessor insn.
20157         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
20158         calls.
20159         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
20160         (nvptx_process_pars): Propagate frames for calls.
20162 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
20164         PR target/85469
20165         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
20166         Removed.
20167         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
20168         (ix86_handle_option): Don't handle OPT_mibt.
20169         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
20170         __SHSTK__.
20171         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
20172         has_ibt and ibt.
20173         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
20174         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
20175         (ix86_target_macros): Define __CET__ with flag_cf_protection
20176         for -fcf-protection.
20177         * config/i386/i386.c (isa2_opts): Remove -mibt.
20178         * config/i386/i386.h (TARGET_IBT): Removed.
20179         (TARGET_IBT_P): Likewise.
20180         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
20181         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
20182         * config/i386/i386.opt (mcet): Update help message.
20183         (mshstk): Likewise.
20184         (mibt): Removed.
20185         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
20186         -mcet as an alias for -mshstk.
20188 2018-04-20  Richard Biener <rguenther@suse.de>
20190         PR middle-end/85475
20191         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
20192         complexity by forcing a single use of the multiply operand.
20194 2018-04-20  Martin Jambor  <mjambor@suse.cz>
20196         ipa/85449
20197         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
20198         recursion dependency to only apply to non-clones.
20200 2018-04-20  Martin Jambor  <mjambor@suse.cz>
20202         ipa/85447
20203         * ipa-cp.c (create_specialized_node): Check that clones of
20204         self-recursive edges exist during IPA-CP.
20206 2018-04-19  Toon Moene  <toon@moene.org>
20208         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
20209         by -O3.
20211 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
20213         PR tree-optimization/85467
20214         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
20215         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
20216         VECTOR_CST element to type.
20218 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
20220         PR target/85397
20221         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
20222         * config/i386/i386.md (builtin_setjmp_setup): Removed.
20223         (builtin_longjmp): Likewise.
20224         (save_stack_nonlocal): New pattern.
20225         (restore_stack_nonlocal): Likewise.
20227 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
20229         PR target/85404
20230         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
20231         Replace ASM_OUTPUT_LABEL with fprintf.
20233 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
20235         PR target/85417
20236         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
20237         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
20238         * config/i386/i386-c.c (ix86_target_macros_internal): Also
20239         define __IBT__ and __SHSTK__ for -fcf-protection.
20240         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
20241         TARGET_IBT.
20242         (ix86_trampoline_init): Likewise.
20243         (x86_output_mi_thunk): Likewise.
20244         (ix86_notrack_prefixed_insn_p): Likewise.
20245         (ix86_option_override_internal): Don't disallow -fcf-protection.
20246         * config/i386/i386.md (rdssp<mode>): Also enable for
20247         -fcf-protection.
20248         (incssp<mode>): Likewise.
20249         (nop_endbr): Likewise.
20250         * config/i386/i386.opt (mcet): Change help message to built-in
20251         functions only.
20252         (mibt): Likewise.
20253         (mshstk): Likewise.
20254         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
20255         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
20256         enable CET built-in functions.
20258 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
20260         * common/config/i386/i386-common.c
20261         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
20262         OPTION_MASK_ISA_MOVDIRI_UNSET,
20263         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
20264         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
20265         * config.gcc (movdirintrin.h): New header.
20266         * config/i386/cpuid.h (bit_MOVDIRI,
20267         bit_MOVDIR64B): New bits.
20268         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
20269         and -mmvodir64b.
20270         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
20271         (VOID, PVOID, PCVOID)): New function types.
20272         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
20273         __builtin_ia32_directstoreu_u64,
20274         __builtin_ia32_movdir64b): New builtins.
20275         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
20276         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
20277         and -mmovdiri.
20278         (ix86_valid_target_attribute_inner_p): Ditto.
20279         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
20280         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
20281         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
20282         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
20283         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
20284         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
20285         (movdiri<mode>, movdir64b_<mode>): New.
20286         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
20287         * config/i386/immintrin.h: Include movdirintrin.h.
20288         * config/i386/movdirintrin.h: New file.
20289         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
20291 2018-04-19  Richard Biener  <rguenther@suse.de>
20293         PR middle-end/85455
20294         * cfg.c (clear_bb_flags): When loop state says we have
20295         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
20297 2018-04-19  Richard Biener  <rguenther@suse.de>
20299         PR tree-optimization/84737
20300         * tree-vect-data-refs.c (vect_copy_ref_info): New function
20301         copying restrict info.
20302         (vect_setup_realignment): Use it.
20303         * tree-vectorizer.h (vect_copy_ref_info): Declare.
20304         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
20305         the first DR to all generated stores.
20306         (vectorizable_load): Likewise for loads.
20308 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
20310         PR tree-optimization/85446
20311         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
20312         the integral and pointer types to have the same precision.
20314         * doc/install.texi: Document --disable-cet being the default and
20315         --enable-cet=auto.
20317 2018-04-18  Martin Liska  <mliska@suse.cz>
20319         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
20320         style.
20322 2018-04-18  Martin Liska  <mliska@suse.cz>
20324         Revert
20325         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
20327         PR ipa/83983
20328         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
20329         arguments if they are comparable.
20331 2018-04-18  Martin Liska  <mliska@suse.cz>
20333         Revert
20334         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
20336         PR lto/84805
20337         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
20338         incomplete types.
20340 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
20342         PR target/85388
20343         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
20344         ENDBR after calling __morestack.
20346 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
20348         PR jit/85384
20349         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
20350         by using gcc_base_ver to generate a gcc_driver_version, and use
20351         it when generating GCC_DRIVER_NAME.
20352         * configure: Regenerate.
20354 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
20356         PR target/81084
20357         * config.gcc: Obsolete powerpc*-*-*spe*.
20359 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
20361         PR debug/84637
20362         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
20363         (stabstr_D): Change type of unum from unsigned int to
20364         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
20365         type.
20367 2018-04-17  Jim Wilson  <jimw@sifive.com>
20369         PR 84856
20370         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
20371         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
20372         Set arg_pointer_offset after using pretend_args_size.
20374 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
20376         PR rtl-optimization/85431
20377         * dse.c (record_store): Ignore zero width stores.
20379         PR sanitizer/85230
20380         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
20381         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
20382         __builtin_stack_restore rather than after it.
20383         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
20384         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
20385         argument instead of virtual_dynamic_stack_rtx.
20387 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20389         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
20390         New prototype.
20391         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20392         Add note to error message to explain internal mapping of overloaded
20393         built-in function name to non-overloaded built-in function name.
20394         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
20395         function.
20397 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
20399         PR target/85424
20400         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
20401         where the inputs overlap with the output.
20403 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
20405         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
20406         (=v, v) alternative and explicit "memory" attribute.
20407         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
20408         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
20409         attributes.
20410         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
20411         "sselog1" type instead of "sselog".
20412         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
20413         "sselog".  Remove explicit "memory" attribute.
20414         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
20415         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
20416         attributes.
20417         (vec_extract_hi_v32hi): Merge all alternatives into one, use
20418         "sselog1" type instead of "sselog".  Remove explicit "memory"
20419         attribute.
20420         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
20421         use "sselog1" type instead of "sselog".  Remove explicit "memory"
20422         attribute.
20423         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
20424         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
20425         attributes.
20426         (vec_extract_hi_v64qi): Merge all alternatives into one, use
20427         "sselog1" type instead of "sselog".  Remove explicit "memory"
20428         attribute.
20429         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
20430         use "sselog1" type instead of "sselog".  Remove explicit "memory"
20431         attribute.
20433         PR target/85430
20434         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
20436         PR middle-end/85414
20437         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
20438         on a SUBREG.
20440 2018-04-17  Martin Jambor  <mjambor@suse.cz>
20442         PR ipa/85421
20443         * ipa-cp.c (create_specialized_node): Call
20444         expand_all_artificial_thunks if necessary.
20446 2018-04-17  Martin Liska  <mliska@suse.cz>
20448         PR lto/85405
20449         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
20450         in message, remote space in between '_G' and '('.
20452 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
20454         PR target/85281
20455         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
20456         avx512f_vmcmp<mode>3<round_saeonly_name>,
20457         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
20458         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
20459         avx512f_rndscale<mode><round_saeonly_name>,
20460         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
20461         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
20462         Use %<iptr>2 instead of %2 for -masm=intel.
20463         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
20464         avx512f_vcvttss2usi<round_saeonly_name>,
20465         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
20466         -masm=intel.
20467         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
20468         avx512f_vcvttsd2usi<round_saeonly_name>,
20469         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
20470         Use %q1 instead of %1 for -masm=intel.
20471         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
20472         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
20473         of %3 for -masm=intel.
20474         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
20475         -masm=intel.
20476         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
20477         -masm=intel.
20478         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
20479         -masm=intel.
20480         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
20481         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
20482         %g1.
20483         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
20484         -masm=intel.
20485         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
20486         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
20487         %g1 and one with %0 and %1.
20488         (avx512er_vmrcp28<mode><round_saeonly_name>,
20489         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
20490         %1 for -masm=intel.
20491         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
20492         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
20493         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
20494         of %0 and %{%4%} for -masm=intel.
20495         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
20496         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
20497         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
20498         order of %0 and %{%5%}%{z%} for -masm=intel.
20500 2018-04-17  Jan Hubicka  <jh@suse.cz>
20502         PR lto/85405
20503         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
20505 2018-04-17  Martin Liska  <mliska@suse.cz>
20507         PR ipa/85329
20508         * multiple_target.c (create_dispatcher_calls): Set apostrophes
20509         for target_clone error message.  Make default implementation
20510         clone to be a local declaration.
20511         (separate_attrs): Add new argument and check for an empty
20512         string.
20513         (expand_target_clones): Handle it.
20514         (ipa_target_clone): Make redirection just for target_clones
20515         functions.
20517 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
20518             Tom de Vries  <tom@codesourcery.com>
20520         PR middle-end/84955
20521         * omp-expand.c (expand_oacc_for): Add dummy false branch for
20522         tiled basic blocks without omp continue statements.
20524 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
20526         PR target/83660
20527         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
20528         vec_extract expression as having side effects to make sure it gets
20529         a cleanup point.
20531 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
20533         PR target/85403
20534         * config/i386/i386.c (get_builtin_code_for_version): Check
20535         error_mark_node.
20537 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
20539         PR target/84331
20540         * config.gcc: Support "skylake".
20541         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20542         PROCESSOR_SKYLAKE.
20543         * config/i386/i386.c (m_SKYLAKE): Define.
20544         (processor_target_table): Add "skylake".
20545         (ix86_option_override_internal): Add "skylake".
20546         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
20547         PROCESSOR_CANNONLAKE.
20548         (get_builtin_code_for_version): Fix priority for
20549         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
20550         PROCESSOR_SKYLAKE-AVX512.
20551         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
20552         (processor_type): Add PROCESSOR_SKYLAKE.
20554 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
20555             Jason Merrill  <jason@redhat.com>
20557         PR c++/85112
20558         * convert.c (convert_to_integer_1): Use direct recursion for
20559         enumeral types and types with a precision less than the number
20560         of bits in their mode.
20562 2018-04-16  Julia Koval  <julia.koval@intel.com>
20564         PR target/84413
20565         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
20566         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
20568 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
20570         PR target/85293
20571         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
20572         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
20573         and -mno-direct-move.
20575 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
20577         PR target/83402
20578         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
20579         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
20580         Ensure negative shifts result in {0}.
20582 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
20584         PR rtl-optimization/79916
20585         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
20586         regs (if any) to define how to gnerate SD moves when LRA is in
20587         progress.
20589 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
20591         PR rtl-optimization/85393
20592         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
20593         * except.c (expand_dw2_landing_pad_for_region): Make static.
20594         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
20595         a label and unconditional jump to old_bb, rather than
20596         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
20597         basic block.
20599         PR rtl-optimization/85376
20600         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
20601         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
20602         instead of a specific value.
20604 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
20605             Bin Cheng  <bin.cheng@arm.com>
20607         PR tree-optimization/82965
20608         PR tree-optimization/83991
20609         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
20610         by_profile_only parameter.
20611         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
20612         information if the loop was predicted to iterate too many times.
20613         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
20615 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
20617         PR lto/71991
20618         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
20619         always inline.
20621 2018-04-13  Martin Liska  <mliska@suse.cz>
20622             Jakub Jelinek  <jakub@redhat.com>
20624         PR middle-end/81657
20625         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
20626         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
20627         * builtins.c (expand_builtin_memory_copy_args): Use
20628         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
20629         handle dest_addr == pc_rtx.
20631 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
20633         PR target/85291
20634         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
20635         asked to not generate direct moves.
20636         (fix_trunc<mode>si2_stfiwx): Similar.
20637         (fix_trunc<mode>si2_internal): Similar.
20639 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
20641         PR debug/83157
20642         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
20643         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
20644         lookup if dest in some wider mode is known to be const0_rtx and
20645         if so, record permanent equivalence for it to be ZERO_EXTEND of
20646         the narrower mode destination.
20648 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
20650         * lto-streamer-out.c (output_function): Revert 259346.
20651         * omp-expand.c (expand_oacc_for): Likewise.
20653 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
20655         PR rtl-optimization/85354
20656         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
20657         * sel-sched.c (sel_global_init): ... here.
20659 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
20661         PR target/85238
20662         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
20663         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
20664         mode for PE-COFF targets.
20665         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
20666         (i386_pe_asm_lto_end): Likewise.
20667         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
20668         (TARGET_ASM_LTO_END): Likewise.
20669         * config/i386/winnt.c (saved_debug_info_level): New static variable.
20670         (i386_pe_asm_lto_start): New function.
20671         (i386_pe_asm_lto_end): Likewise.
20673 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
20674             Richard Biener  <rguenther@suse.de>
20676         PR middle-end/84955
20677         * lto-streamer-out.c (output_function): Fix CFG loop state before
20678         streaming out.
20679         * omp-expand.c (expand_oacc_for): Handle calls to internal
20680         functions like regular functions.
20682 2018-04-12  Richard Biener  <rguenther@suse.de>
20684         PR lto/85371
20685         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
20686         for the early LTO debug to properly generate references to it
20687         during DIE emission.  Do not re-use that for the skeleton for
20688         split-dwarf.
20689         (dwarf2out_early_finish): Likewise.
20691 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
20693         PR target/85328
20694         * config/i386/sse.md
20695         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
20696         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
20697         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
20698         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
20699         and output is a reg, avoid creating invalid lowpart subreg, but
20700         instead split into a 512-bit move.  Don't split if not AVX512VL,
20701         input is xmm16+ reg and output is a mem.
20702         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
20703         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
20704         xmm16+ reg and output is a mem.
20706 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
20708         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
20709         also for flag_dwarf2_cfi_asm.
20711 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
20713         PR rtl-optimization/85342
20714         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
20715         a bool scalar var inside of the loop instead.  Don't try to update
20716         recog_data.operand after failed apply_change_group.
20718 2018-04-12  Tom de Vries  <tom@codesourcery.com>
20720         PR target/85296
20721         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
20722         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
20723         array with flexible array member as array without given dimension.
20724         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
20725         argument for undefined param to true.
20727 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
20729         PR target/85321
20730         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
20731         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
20732         from PowerPC section.
20733         * config/rs6000/sysv4.opt (mcall-): Improve help text.
20734         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
20735         help text that is too long.
20736         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
20737         help text that is too long.
20738         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
20739         help text that is too long.
20741 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
20743         * config/alpha/alpha.md (stack_probe_internal): Rename
20744         from "probe_stack".  Update all callers.
20746 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
20748         PR rtl-optimization/84566
20749         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
20750         sched_macro_fuse_insns.
20752 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
20754         PR target/84301
20755         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
20756         (compute_block_dependences): ... from here.
20758 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
20760         PR tree-optimization/85331
20761         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
20762         from int to HOST_WIDE_INT.
20764 2018-04-11  Martin Jambor  <mjambor@suse.cz>
20766         PR ipa/84149
20767         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
20768         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
20769         not the same as the source val.
20770         (cgraph_edge_brings_value_p): New parameter.
20771         (gather_edges_for_value): Pass destination value to
20772         cgraph_edge_brings_value_p.
20773         (perhaps_add_new_callers): Likewise.
20774         (get_info_about_necessary_edges): Likewise and exclude values brought
20775         only by self-recursive edges.
20776         (create_specialized_node): Redirect only clones of self-calling edges.
20777         (+self_recursive_pass_through_p): New function.
20778         (find_more_scalar_values_for_callers_subset): Use it.
20779         (find_aggregate_values_for_callers_subset): Likewise.
20780         (known_aggs_to_agg_replacement_list): Removed.
20781         (decide_whether_version_node): Re-calculate known constants for all
20782         remaining context clones.
20784 2018-04-11  Richard Biener  <rguenther@suse.de>
20786         PR lto/85339
20787         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
20788         from early DWARF output.
20789         (dwarf2out_early_finish): Output line info unconditionally into
20790         early DWARF and add reference to it.
20792 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
20794         PR target/85281
20795         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
20796         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
20797         other than V2DFmode using iptr mode attribute.
20798         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
20800 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
20802         PR rtl-optimization/84659
20803         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
20805 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
20807         PR debug/85302
20808         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
20809         SIZEP is NULL.
20810         (output_loc_list): Pass address of a dummy size variable even in the
20811         locview handling loop.
20812         (index_location_lists): Add comment on why skip_loc_list_entry can't
20813         call size_of_locs.
20815 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20817         PR target/85261
20818         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
20819         into register.
20821 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
20823         PR target/85321
20824         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
20825         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
20826         and -mstring-compare-inline-limit.
20828 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
20830         PR target/85287
20831         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
20832         for stack clash protection in a register whenever we need it to be in
20833         a register.
20835 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
20837         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
20838         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
20840 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
20842         PR target/85321
20843         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
20844         the help text.
20845         (mlong-double-): Ditto.
20846         * config/rs6000/sysv4.opt (msdata=): Ditto.
20847         (mtls-size=): Ditto.
20849 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20851         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
20852         erroneous entries for
20853         "vector int vec_ldl (int, long int *)", and
20854         "vector unsigned int vec_ldl (int, unsigned long int *)".
20855         Add comments and entries for
20856         "vector bool char vec_ldl (int, bool char *)",
20857         "vector bool short vec_ldl (int, bool short *)",
20858         "vector bool int vec_ldl (int, bool int *)",
20859         "vector bool long long vec_ldl (int, bool long long *)",
20860         "vector pixel vec_ldl (int, pixel *)",
20861         "vector long long vec_ldl (int, long long *)",
20862         "vector unsigned long long vec_ldl (int, unsigned long long *)".
20863         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
20864         type tree bool_long_long_type_node and correct definition of
20865         bool_V2DI_type_node to make reference to this new type tree.
20866         (rs6000_mangle_type): Replace erroneous reference to
20867         bool_long_type_node with bool_long_long_type_node.
20868         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
20869         comments to emphasize sign distinctions for char and int types and
20870         replace RS6000_BTI_bool_long constant with
20871         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
20872         use of RS6000_BTI_pixel.
20873         (bool_long_type_node): Remove this macro definition.
20874         (bool_long_long_type_node): New macro definition
20876 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
20878         PR rtl-optimization/85300
20879         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
20880         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
20881         simplify_unary_operation fails.
20883 2018-04-10  Martin Liska  <mliska@suse.cz>
20885         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
20886         cgraph_edge and ipa_ref.
20888 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
20890         PR target/85177
20891         PR target/85255
20892         * config/i386/sse.md
20893         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
20894         computation of the VEC_MERGE selector from mask.
20895         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
20896         Fix decoding of the VEC_MERGE selector into mask.
20898 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
20900         PR tree-optimization/85286
20901         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
20903 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
20905         * final.c (final_1): Set insn_last_address as well as
20906         insn_current_address.
20908 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20910         PR target/85173
20911         * explow.c (emit_stack_probe): Call validize_mem on memory location
20912         before passing it to gen_probe_stack.  Create address operand and
20913         legitimize it for the probe_stack_address case.
20915 2018-04-09  Jan Hubicka  <jh@suse.cz>
20917         PR lto/85078
20918         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
20919         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
20920         * tree.c (free_lang_data_in_type): Fix handling of binfos;
20921         walk basetypes.
20922         (free_lang_data): Rebuild type inheritance graph.
20924 2018-04-09  Martin Sebor  <msebor@redhat.com>
20926         * invoke.texi (-finline-small-functions): Mention other optimization
20927         options.
20928         (-findirect-inlining, -fpartial-inlining): Same.
20929         (-finline-functions-called-once): Same.
20930         (-freorder-blocks-and-partition): Same.
20932 2018-04-09  Jan Hubicka  <jh@suse.cz>
20934         PR rtl/84058
20935         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
20936         jumps; choose last target that matches the criteria (i.e.
20937         no partition changes for non-crossing jumps).
20938         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
20939         support for redirecting crossing jumps to non-crossing.
20941 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
20943         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
20944         also for naked functions.
20946 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
20948         * config/arc/arc.md (add_shift): New pattern.
20949         (add_shift2): Likewise.
20950         (sub_shift): Likewise.
20951         (sub_shift_cmp0_noout): Likewise.
20952         (compare_si_ashiftsi): Likewise.
20953         (xbfu_cmp0_noout): New combine pattern.
20954         (xbfu_cmp0"): Likewise.
20955         (movsi_set_cc_insn): Place the predicable variant first.
20956         (commutative_binary_cmp0_noout): Remove clobber.
20957         (commutative_binary_cmp0): New pattern.
20958         (noncommutative_binary_cmp0): Likewise.
20959         (noncommutative_binary_cmp0_noout): Likewise.
20960         (noncommutative_binary_comparison_result_used): Removed.
20961         (rsub_cmp0): New pattern.
20962         (rsub_cmp0_noout): Likewise.
20963         (extzvsi): Changed, keep only meaningful variants.
20964         (SQH, SEZ): New iterators.
20965         (SQH_postfix): New mode attribute.
20966         (SEZ_prefix): New code attribute.
20967         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
20968         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
20969         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
20970         of numerical value.
20971         (noncommutative_operator): Check the availability of barrel
20972         shifter option.
20974 2018-04-09  Richard Biener  <rguenther@suse.de>
20976         PR tree-optimization/85284
20977         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
20978         Only use the niter constraining form of simple_iv when the exit
20979         is always executed.
20981 2018-04-09  Tom de Vries  <tom@codesourcery.com>
20983         PR target/84041
20984         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
20985         (define_expand "*memory_barrier"): New define_expand.
20986         (define_insn "memory_barrier"): New insn.
20988 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
20990         PR rtl-optimization/80463
20991         PR rtl-optimization/83972
20992         PR rtl-optimization/83480
20994         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
20995         correct producer for the insn.
20996         (tidy_control_flow): Fixup seqnos in case of debug insns.
20998 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21000         PR rtl-optimization/83913
21002         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
21003         different sched-times when merging exprs.
21005 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21007         PR rtl-optimization/83962
21009         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
21010         tidy_fallthru_edge and tidy_control_flow.
21012 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
21014         PR rtl-optimization/83530
21016         * sel-sched.c (force_next_insn): New global variable.
21017         (remove_insn_for_debug): When force_next_insn is true, also leave only
21018         next insn in the ready list.
21019         (sel_sched_region): When the region wasn't scheduled, make another pass
21020         over it with force_next_insn set to 1.
21022 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
21024         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
21025         into tm_file.
21026         * config/nds32/constants.md (unspec_volatile_element): Add enum values
21027         for interrupt control.
21028         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
21029         functions for interrupt control.
21030         * config/nds32/nds32-intrinsic.md: Likewise.
21031         * config/nds32/nds32_intrinsic.h: Likewise.
21032         * config/nds32/nds32.h (nds32_builtins): Likewise.
21034 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
21036         * config/nds32/nds32.c (nds32_init_machine_status,
21037         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
21038         strict_aligned_p field.
21039         (nds32_expand_to_rtl_hook): New function.
21040         (TARGET_EXPAND_TO_RTL_HOOK): Define.
21041         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
21043 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21044             Chung-Ju Wu  <jasonwucj@gmail.com>
21046         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
21047         * config/nds32/nds32-n7.md: New file.
21048         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
21049         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
21050         pipeline.
21051         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
21052         * config/nds32/nds32.md (pipeline_model): Add n7.
21053         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
21054         * config/nds32/pipelines.md: Include n7 settings.
21056 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21057             Chung-Ju Wu  <jasonwucj@gmail.com>
21059         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
21060         * config/nds32/nds32-e8.md: New file.
21061         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
21062         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
21063         pipeline.
21064         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
21065         * config/nds32/nds32.md (pipeline_model): Add e8.
21066         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
21067         * config/nds32/pipelines.md: Include e8 settings.
21069 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21070             Chung-Ju Wu  <jasonwucj@gmail.com>
21072         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
21073         * config/nds32/nds32-n8.md: New file.
21074         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
21075         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
21076         pipeline.
21077         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
21078         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
21079         * config/nds32/nds32.md (pipeline_model): Add n8.
21080         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
21081         * config/nds32/pipelines.md: Include n8 settings.
21083 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
21084             Chung-Ju Wu  <jasonwucj@gmail.com>
21086         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
21087         * config/nds32/nds32-n9-2r1w.md: New file.
21088         * config/nds32/nds32-n9-3r2w.md: New file.
21089         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
21090         nds32_register_ports): New or modify for cpu n9.
21091         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
21092         pipeline.
21093         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
21094         * config/nds32/nds32-utils.c: New file.
21095         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
21096         TARGET_MUL_SLOW): Define.
21097         * config/nds32/nds32.md (pipeline_model): New attribute.
21098         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
21099         New options that support cpu n9.
21100         * config/nds32/pipelines.md: Include n9 settings.
21101         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
21103 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
21105         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
21106         information if necessary.
21107         (output_cond_branch_compare_zero): Likewise.
21108         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
21109         (nds32_target_alignment): Refine for alignment.
21110         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
21111         (FUNCTION_BOUNDARY): Modify.
21112         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
21113         align case.
21114         * config/nds32/nds32.opt (malways-align, malign-functions): New.
21116 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
21118         * config/nds32/constants.md (unspec_volatile_element): Add values for
21119         TLB operation and data prefetch.
21120         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
21121         functions for TLB operation and data prefetch.
21122         * config/nds32/nds32-intrinsic.md: Likewise.
21123         * config/nds32/nds32_intrinsic.h: Likewise.
21124         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
21125         (nds32_print_operand): Likewise.
21126         * config/nds32/nds32.h (nds32_builtins): Likewise.
21128 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
21129         Andrew Pinski <pinsika@gcc.gnu.org>
21131         PR middle-end/82976
21132         * match.pd: Use constant_boolean_node of correct type instead of
21133         boolean_true_node or boolean_false_node for simplifying
21134         pointer comparisons to zero.
21136 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
21138         PR tree-optimization/80021
21139         * tree.c (verify_type_variant): Make error call in verify_variant_match
21140         translatable and remove final full stop.
21142 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21144         * config/nds32/constants.md (unspec_volatile_element): Add
21145         UNSPEC_VOLATILE_EH_RETURN.
21146         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
21147         nds32_output_stack_pop): Support dwarf exception handling process.
21148         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
21149         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
21150         exception handling process.
21151         (nds32_compute_stack_frame): Likewise.
21152         (nds32_return_addr_rtx): Likewise.
21153         (nds32_initial_elimination_offset): Likewise.
21154         (nds32_expand_prologue): Likewise.
21155         (nds32_expand_epilogue): Likewise.
21156         (nds32_dynamic_chain_address): New function.
21157         * config/nds32/nds32.h (machine_function): Add fields for dwarf
21158         exception handling.
21159         (DYNAMIC_CHAIN_ADDRESS): Define.
21160         (EH_RETURN_DATA_REGNO): Define.
21161         (EH_RETURN_STACKADJ_RTX): Define.
21162         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
21163         patterns for dwarf exception handling.
21165 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21167         * config/nds32/nds32.h: Clean up obsolete macros.
21169 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21171         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
21172         Add enum values for particular instructions.
21173         * config/nds32/nds32-intrinsic.c: Implementation of expanding
21174         particular intrinsic functions.
21175         * config/nds32/nds32-intrinsic.md: Likewise.
21176         * config/nds32/nds32_intrinsic.h: Likewise.
21177         * config/nds32/nds32.h (nds32_builtins): Likewise.
21178         * config/nds32/nds32.md (type): Add pbsad and pbsada.
21179         (btst, ave): New patterns for particular instructions.
21181 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21183         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
21184         Add enum values for atomic load/store and memory sync.
21185         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
21186         and memory sync.
21187         * config/nds32/nds32-intrinsic.md: Likewise.
21188         * config/nds32/nds32_intrinsic.h: Likewise.
21189         * config/nds32/nds32.h (nds32_builtins): Likewise.
21191 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
21193         PR tree-optimization/85257
21194         * fold-const.c (native_encode_vector): If not all elts could fit
21195         and off is -1, return 0 rather than offset.
21196         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
21197         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
21198         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
21199         adjust buffer in native_interpret_expr call.
21201 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21203         * config/nds32/constants.md (unspec_volatile_element): Add cache
21204         control enum values.
21205         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
21206         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
21207         * config/nds32/nds32.c (nds32_cctl_names): New.
21208         (nds32_print_operand): Handle cache control register names.
21209         * config/nds32/nds32.h (nds32_builtins): New enum values.
21210         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
21211         macros.
21212         * config/nds32/nds32.md (type): Add mmu.
21213         * config/nds32/pipelines.md (simple_insn): Add mmu.
21215 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21217         * config/nds32/nds32.md (type): Remove call.
21218         * config/nds32/pipelines.md (simple_insn): Likewise.
21220 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21222         * config/nds32/constants.md (unspec_volatile_element): Add
21223         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
21224         UNSPEC_VOLATILE_FMFCFG.
21225         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
21226         description for fmfcfg and fmfcsr.
21227         (bdesc_1arg): Add fmtcsr.
21228         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
21229         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
21230         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
21231         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
21232         unspec_fmfcfg): New patterns.
21233         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
21234         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
21235         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
21236         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
21237         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
21238         __nds32__fmfcfg): Define.
21240 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
21242         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
21243         intrinsic register names.
21244         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
21245         intrinsic register enum values and macros.
21247 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
21249         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
21250         for load/store addressing form.
21251         (nds32_print_operand_address): Likewise.
21253 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
21255         PR target/85196
21256         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
21257         based on LABEL_REF.  Remove useless assertion.
21258         (pic_address_needs_scratch): Fix formatting.
21259         (sparc_legitimize_pic_address): Minor tweaks.
21260         (sparc_delegitimize_address): Adjust assertion accordingly.
21261         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
21262         into symbolic_operand.
21263         (movsi_high_pic_label_ref): Likewise.
21264         (movsi_lo_sum_pic_label_ref): Likewise.
21265         (movdi_pic_label_ref): Likewise.
21266         (movdi_high_pic_label_ref): Likewise.
21267         (movdi_lo_sum_pic_label_ref): Likewise.
21269 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
21271         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
21272         custom LIB_SPEC setup.
21274 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
21275             Kito Cheng  <kito.cheng@gmail.com>
21277         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
21278         * config/riscv/freebsd.h: New.
21280 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
21282         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
21283         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
21284         file.
21286 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
21287             Kito Cheng  <kito.cheng@gmail.com>
21289         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
21290         nds32_output_call, nds32_symbol_binds_local_p): New functions.
21291         * config/nds32/nds32-protos.h (nds32_output_call,
21292         nds32_output_return): Declare.
21293         * config/nds32/nds32.md: Refine all the call and return patterns.
21295 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
21297         PR debug/85252
21298         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
21299         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
21301         PR rtl-optimization/84872
21302         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
21303         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
21304         EDGE_CROSSING edge.
21306 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
21308         * expr.c (copy_blkmode_to_reg): Revert 254862.
21309         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
21311 2018-04-06  Richard Biener  <rguenther@suse.de>
21313         PR middle-end/85244
21314         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
21315         after seeing a component reference with an adjacent field.  Treat
21316         refs to arrays at struct end of external decls similar to
21317         refs to unconstrained commons.
21319 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
21321         PR sanitizer/85213
21322         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
21323         look through SAVE_EXPRs with non-side-effects argument.  Adjust
21324         recursive calls.
21325         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
21326         save_p here.
21328 2018-04-06  Richard Biener  <rguenther@suse.de>
21330         PR middle-end/85180
21331         * alias.c (find_base_term): New wrapper around find_base_term
21332         unwinding CSELIB_VAL_PTR changes.
21333         (find_base_term): Do not restore CSELIB_VAL_PTR during the
21334         recursion.
21336 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21338         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
21339         instructions.
21340         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
21341         constant definitions.
21342         ("nop"): lr 0,0 -> nopr r0
21343         ("nop_lr0", "nop_lr1"): New insn definitions.
21345 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
21347         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
21348         NDS32_V3PUSH_AVAILABLE_P macro.
21350 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
21351             Chung-Ju Wu  <jasonwucj@gmail.com>
21353         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
21354         (nds32*-*-*): Add float and fpu_config into supported_defaults.
21355         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
21356         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
21357         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
21358         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
21359         * config/nds32/constraints.md: New constraints and checking for hard
21360         float configuration.
21361         * config/nds32/iterators.md: New mode iterator and attribute for hard
21362         float configuration.
21363         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
21364         patterns.
21365         * config/nds32/nds32-fpu.md: New file.
21366         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
21367         deal with hard float code generation.
21368         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
21369         ARCH_V3S.
21370         (abi_type, float_reg_number): New enum type.
21371         * config/nds32/nds32-predicates.c: New predicates for hard float.
21372         * config/nds32/nds32-protos.h: Declare functions for hard float.
21373         * config/nds32/nds32.c: Implementation for hard float configuration.
21374         * config/nds32/nds32.h: Definitions for hard float configuration.
21375         * config/nds32/nds32.md: Include hard float machine description and
21376         modify patterns for hard float configuration.
21377         * config/nds32/nds32.opt: New options for hard float configuration.
21378         * config/nds32/predicates.md: New predicates for hard float
21379         configuration.
21381 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
21383         * common/config/nds32/nds32-common.c
21384         (nds32_option_optimization_table): Enable -mreleax-hint by default.
21386 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
21388         PR middle-end/85195
21389         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
21390         CONSTRUCTOR_ELT (ctor, ...)->value.
21392 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
21394         PR target/85193
21395         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
21397 2018-04-05  Tom de Vries  <tom@codesourcery.com>
21399         PR target/85204
21400         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
21401         cond jump.
21403 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
21404             Kito Cheng  <kito.cheng@gmail.com>
21406         * config/nds32/constraints.md (U33): Fine-tune checking condition.
21407         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
21408         * config/nds32/nds32.h (nds32_16bit_address_type): Add
21409         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
21411 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
21412             Kito Cheng  <kito.cheng@gmail.com>
21414         * config/nds32/constraints.md (Ufe): New memory constraint.
21415         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
21416         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
21417         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
21418         operands.
21419         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
21420         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
21422 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21424         * config/nds32/nds32.md: Use optimize_size in the condition for
21425         alu-shift instructions.
21427 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21429         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
21431 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21433         * config/nds32/nds32.md (negsi2): Refine pattern.
21435 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
21436             Chung-Ju Wu  <jasonwucj@gmail.com>
21438         * config/nds32/iterators.md (shift_rotate): New code iterator.
21439         (shift): New code attribute.
21440         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
21441         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
21442         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
21443         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
21444         bit-wise operations.
21445         (andsi3, *andsi3): Ditto.
21446         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
21447         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
21448         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
21449         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
21450         nds32_ior_operand, nds32_xor_operand): New predicates.
21452 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21454         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
21455         (addsi3, subsi3): ... this.
21457 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21459         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
21461 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21463         * config/nds32/nds32.md: Adjust indention.
21465 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
21467         * config/nds32/nds32.md (feature): New attribute.
21469 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
21471         * config/nds32/nds32.md (subtype): New attribute.
21473 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
21475         PR target/85203
21476         * config/arm/arm-builtins.c (arm_expand_builtin): Change
21477         expansion to perform a bitwise AND of the argument followed by a
21478         boolean negation of the result.
21480 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
21482         PR rtl-optimization/84878
21483         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
21484         the basic block.  Assert the use reference is not artificial and that
21485         it has an associated insn.
21487 2018-04-04  Michael Matz  <matz@suse.de>
21489         * builtins.c (compute_objsize): Pass correct operand
21490         to array_at_struct_end_p.
21492 2018-04-04  Richard Biener  <rguenther@suse.de>
21494         PR lto/85176
21495         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
21496         from contexts for DINFO_LEVEL_TERSE and below.
21498 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
21500         * config/nds32/nds32-doubleword.md (move_<mode>): Require
21501         resiter_operand condition.
21502         * config/nds32/nds32.md (*move<mode>): Ditto.
21504 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
21505             Monk Chiang  <sh.chiang04@gmail.com>
21507         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
21509 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
21511         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
21513 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
21514             Kito Cheng  <kito.cheng@gmail.com>
21516         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
21517         nds32_cond_code_str, output_cond_branch,
21518         output_cond_branch_compare_zero, nds32_expand_cbranch,
21519         nds32_expand_cstore, nds32_expand_movcc,
21520         nds32_output_cbranchsi4_equality_zero,
21521         nds32_output_cbranchsi4_equality_reg,
21522         nds32_output_cbranchsi4_equality_reg_or_const_int,
21523         nds32_output_cbranchsi4_greater_less_zero: New functions.
21524         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
21525         nds32_expand_cstore, nds32_expand_movcc,
21526         nds32_output_cbranchsi4_equality_zero,
21527         nds32_output_cbranchsi4_equality_reg,
21528         nds32_output_cbranchsi4_equality_reg_or_const_int,
21529         nds32_output_cbranchsi4_greater_less_zero): Declare.
21530         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
21531         nds32_rimm11s_operand): New predicates.
21532         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
21533         * config/nds32/nds32.md: Rewrite all the branch and conditional move
21534         patterns.
21536 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
21538         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
21539         * config/nds32/nds32.md: Ditto.
21540         * config/nds32/pipelines.md: Ditto.
21542 2018-04-04  Richard Biener  <rguenther@suse.de>
21544         PR tree-optimization/85168
21545         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
21546         propagating abnormals.
21548 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
21550         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
21552 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
21553             Kito Cheng  <kito.cheng@gmail.com>
21555         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
21556         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
21557         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
21558         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
21559         * config/nds32/nds32.md (sibcall_internal): New.
21560         (sibcall_register): Remove.
21561         (sibcall_immediate): Remove.
21562         (sibcall_value_internal): New.
21563         (sibcall_value_register): Remove.
21564         (sibcall_value_immediate): Remove.
21565         * config/nds32/predicates.md (nds32_general_register_operand): New.
21566         (nds32_call_address_operand): New.
21568 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
21570         PR rtl-optimization/85167
21571         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
21572         bb_defs if *split_p, instead preinitialize it to NULL.
21574         PR tree-optimization/85156
21575         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
21576         evaluating the argument multiple times.
21578 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
21580         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
21581         than vector.
21582         (_mm_cvtpd_ps): Likewise.
21583         (_mm_cvttpd_epi32): Likewise.
21584         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
21585         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
21586         vector, pixel, and bool following altivec.h include.
21588 2018-04-03  Martin Sebor  <msebor@redhat.com>
21590         * doc/extend.texi (Common Function Attributes): Clarify.
21591         (const attribute): Likewise.
21592         (pure attribute): Likewise.
21594 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
21596         PR target/85169
21597         * config/i386/i386.c (ix86_expand_vector_set): Use
21598         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
21600 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
21602         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
21603         instructions when changing rounding bits to preserve precision bits
21604         in the x87 control word.
21606 2018-04-03  Martin Liska  <mliska@suse.cz>
21608         PR tree-optimization/82491
21609         * rtl.h (strip_offset_and_add): Replace += suboffset with
21610         poly_uint64 () + suboffset.
21612 2018-03-29  Martin Liska  <mliska@suse.cz>
21613             Martin Jambor  <mjambor@suse.cz>
21615         PR ipa/84947
21616         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
21617         param_type is not an integral or pointer type.
21619 2018-04-03  Richard Biener  <rguenther@suse.de>
21621         * sese.h (recompute_all_dominators): Remove.
21623 2018-04-02  Martin Sebor  <msebor@redhat.com>
21625         * doc/invoke.texi (-Wrestrict): Fix typos.
21627 2018-04-02  Jim Wilson  <jimw@sifive.com>
21629         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
21630         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
21631         (<optab>di3, <optab>si3_extend): Likewise.
21632         (<optab>si3_mask, <optab>si3_mask_1): New.
21633         (<optab>di3_mask, <optab>di3_mask_1): New.
21634         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
21635         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
21636         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
21638 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
21640         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
21641         example.
21643 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
21645         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
21646         (nds32_canonicalize_comparison): New function.
21648 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
21649             Kito Cheng  <kito.cheng@gmail.com>
21650             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
21652         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
21653         * config/nds32/constants.md (unspec_volatile_element): Add
21654         UNSPEC_VOLATILE_RELAX_GROUP.
21655         * config/nds32/nds32-relax-opt.c: New file.
21656         * config/nds32/nds32-predicates.c
21657         (nds32_symbol_load_store_p): New function.
21658         * config/nds32/nds32-protos.h
21659         (nds32_symbol_load_store_p): Declare function.
21660         (make_pass_nds32_relax_opt): Declare new rtl pass function.
21661         * config/nds32/nds32.c
21662         (nds32_register_pass): New function to register pass.
21663         (nds32_register_passes): New function to register passes.
21664         * config/nds32/nds32.md (relax_group): New pattern.
21665         * config/nds32/nds32.opt (mrelax-hint): New option.
21666         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
21668 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
21670         * config/nds32/t-nds32: Modify files dependency.
21672 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
21674         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
21675         (PROFILE_HOOK): Define its implementation.
21677 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
21679         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
21680         type and 32-bit size.
21682 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
21684         PR middle-end/85090
21685         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
21686         (V_128_256): New mode iterator.
21687         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
21688         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
21689         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
21690         of V.
21691         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
21692         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
21694 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
21696         PR target/83315
21697         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
21698         NaN inputs correctly.
21700 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
21702         PR target/80546
21703         * config/rs6000/vsx.md (??r): New mode attribute.
21704         (*vsx_mov<mode>_64bit): Use it.
21705         (*vsx_mov<mode>_32bit): Likewise.
21707 2018-03-30  Martin Sebor  <msebor@redhat.com>
21709         PR tree-optimization/84818
21710         * builtins.c (check_access): Use warning_n.
21712 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
21714         PR target/83822
21715         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
21716         condition.
21717         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
21718         condition.
21720 2018-03-30  Julia Koval  <julia.koval@intel.com>
21722         PR target/84413
21723         * x86-tune.def (movx, partial_reg_dependency): Enable for
21724         m_SKYLAKE_AVX512.
21726 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
21728         PR inline-asm/84985
21729         * lra-constraints.c (process_alt_operands): Move setting
21730         this_alternative_matches below.
21732 2018-03-29  Martin Liska  <mliska@suse.cz>
21734         PR lto/84995.
21735         * doc/invoke.texi: Document how LTO works with debug info.
21736         Describe auto-load support of binutils.  Mention 'x86-64'
21737         as valid option value of -march option.
21739 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
21741         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
21743         PR c/85094
21744         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
21745         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
21746         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
21747         checking.
21749 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
21751         PR target/84912
21752         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
21753         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
21754         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
21755         for RS6000_BTM_POWERPC64.
21756         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
21757         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
21758         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
21759         definition.
21760         (DIVDE): Use it.
21761         (DIVDEU): Likewise.
21763 2018-03-28  Carl Love  <cel@us.ibm.com>
21765         Revert
21766         2017-09-27  Carl Love  <cel@us.ibm.com>
21768         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
21769         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
21770         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
21771         fctiw instruction.
21773 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21775         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
21776         instead of __vector bool.
21777         (_mm_max_pu8): Likewise.
21778         (_mm_min_pi16): Likewise.
21780 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
21782         PR target/84912
21783         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
21784         (DIVWEUO): Likewise.
21785         (DIVDEO): Likewise.
21786         (DIVDEUO): Likewise.
21787         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
21788         DIVWEUO and DIVDEUO.
21789         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
21790         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
21791         (div_extend): Likewise.
21792         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
21793         builtin function.
21794         (__builtin_divweuo): Likewise.
21795         (__builtin_divdeo): Likewise.
21796         (__builtin_divdeuo): Likewise.
21798 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
21800         PR target/85095
21801         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
21802         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
21804         PR tree-optimization/82004
21805         * gimple-match-head.c (optimize_pow_to_exp): New function.
21806         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
21807         Don't fold to exp if optimize_pow_to_exp is false.
21809 2018-03-28  Martin Liska  <mliska@suse.cz>
21811         PR other/84819
21812         * calls.c (initialize_argument_information): Fix trailing space.
21813         * common.opt: Fix typo and provide better explanation for
21814         -fsanitize-coverage option.
21815         * config/i386/i386.opt: Fix typo.
21817 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
21818             Martin Liska  <mliska@suse.cz>
21820         PR sanitizer/85081
21821         * gimplify.c (asan_poison_variable): Don't do the check for
21822         gimplify_omp_ctxp here.
21823         (gimplify_decl_expr): Do it here.
21824         (gimplify_target_expr): Likewise.
21826 2018-03-28  Martin Liska  <mliska@suse.cz>
21828         PR target/84988
21829         * config/i386/i386.c (ix86_function_arg_advance): Do not call
21830         chkp_type_bounds_count if MPX is not enabled.
21832 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
21834         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
21836 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
21838         PR target/84914
21839         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
21840         function to create the function decl for complex long double
21841         multiply and divide for -mabi=ieeelongdouble.
21842         (init_float128_ieee): Call it.
21844 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
21846         PR target/85044
21847         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
21848         -fcf-protection=branch -mibt.
21849         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
21851 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
21853         PR target/81863
21854         * config/arm/arm.c (arm_valid_symbolic_address): Handle
21855         arm_word_relocations.
21857 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
21859         PR target/85056
21860         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
21861         extern array declarations.
21863 2018-03-27  Richard Biener  <rguenther@suse.de>
21865         PR middle-end/84067
21866         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
21867         explicit single_use checks.
21869 2018-03-27  Richard Biener  <rguenther@suse.de>
21871         PR tree-optimization/85082
21872         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
21873         Valueize the VUSE.
21875 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
21877         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
21878         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
21879         Turn on fasynchronous-unwind-tables and funwind-tables.
21881 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
21883         PR target/85073
21884         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
21885         (*bmi_blsr_<mode>_ccz): Ditto.
21887 2018-03-26  Tom de Vries  <tom@codesourcery.com>
21889         PR tree-optimization/85063
21890         * omp-general.c (offloading_function_p): New function.  Factor out
21891         of ...
21892         * omp-offload.c (pass_omp_target_link::gate): ... here.
21893         * omp-general.h (offloading_function_p): Declare.
21894         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
21895         with attribute omp declare target for offloading functions.
21897 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
21899         PR tree-optimization/84005
21900         * tree-data-ref.h (get_base_for_alignment): Declare.
21901         * tree-data-ref.c (get_base_for_alignment_1): New function.
21902         (get_base_for_alignment): Likewise.
21903         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
21904         get_base_for_alignment to find a suitable base object, instead
21905         of always using drb->base_address.
21907 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
21909         PR inline-asm/85022
21910         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
21911         known size by default.
21913 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
21915         PR inline-asm/85030
21916         * lra-constraints.c (process_alt_operands): Don't match BLKmode
21917         and non BLKmode operands.
21919 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21921         PR target/85026
21922         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
21923         Clean up attributes.
21925 2018-03-23  Richard Biener  <rguenther@suse.de>
21927         PR debug/85020
21928         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
21929         we are going to emit early debug for LTO.
21931 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
21933         PR inline-asm/85034
21934         * function.c (match_asm_constraints_1): Don't optimize if input
21935         doesn't satisfy general_operand predicate for output's mode.
21937         PR inline-asm/85022
21938         * alias.c (write_dependence_p): Don't require for x_canonicalized
21939         non-VOIDmode if x has VOIDmode.
21941         PR sanitizer/85029
21942         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
21943         just don't try to optimize it rather than assert it never happens.
21945 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21947         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
21948         macro expansions for definition of ST_INTERNAL_<mode> and
21949         LD_INTERNAL_<mode> builtins.
21950         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
21951         Remove prototype.
21952         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
21953         function.
21954         (altivec_expand_st_builtin): Likewise.
21955         (altivec_expand_builtin): Remove calls to deleted functions.
21956         (rs6000_address_for_altivec): Delete this function.
21957         * config/rs6000/vector.md: Remove expands for
21958         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
21960 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
21962         PR target/84826
21963         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
21964         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
21965         re-computing once computed.
21966         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
21967         (arm_init_machine_status): Initialize
21968         machine->static_chain_stack_bytes.
21970 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21972         PR target/84760
21973         * doc/extend.texi: Add four new prototypes for vec_ld.
21974         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
21975         definitions for more logical presentation.
21976         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
21977         entries for V1TI variants of __builtin_altivec_ld builtin.
21978         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
21979         handling of V1TI variant of LVX icode pattern.
21980         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
21981         (rs6000_gimple_fold_builtin): Likewise.
21982         (altivec_init_builtins): Add code to define
21983         __builtin_altivec_lvx_v1ti function.
21985 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
21987         PR inline-asm/84941
21988         * function.c (match_asm_constraints_1): Don't do the optimization
21989         if input isn't a REG, SUBREG, MEM or constant.
21991 2018-03-22  Tom de Vries  <tom@codesourcery.com>
21993         PR tree-optimization/84956
21994         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
21995         bb_has_abnormal_pred.
21997 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
21999         PR sanitizer/85018
22000         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
22001         DECL_INITIAL (decl) to decl at the end.
22002         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
22003         adjust the comment.
22005 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
22007         * doc/extend.texi (__builtin_tgmath): Document when complex
22008         integer types are treated as _Complex _Float64.
22010 2018-03-21  Tom de Vries  <tom@codesourcery.com>
22012         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
22014 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
22016         PR tree-optimization/84960
22017         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
22018         if it is ENTRY block, move them into single succ of ENTRY in that case.
22020 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
22022         PR tree-optimization/84811
22023         * poly-int.h (poly_span_traits): Remove the T3 parameter and
22024         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
22025         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
22026         (known_subrange_p): Update accordingly.  Cast each value involved
22027         in the size comparison, rather than casting the result of the
22028         subtraction.
22030 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
22032         PR tree-optimization/84982
22033         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
22034         by flipping the least significant bit rather than all bits from
22035         bitpos to bitpos + bitsize - 1.
22037 2018-03-21  Nathan Sidwell  <nathan@acm.org>
22039         * doc/extend.texi (Deprecated Features): Remove mention of
22040         long-deleted deprecations.
22042 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22044         PR jit/84288
22045         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
22046         * configure: Regenerate.
22048 2018-03-21  Tom de Vries  <tom@codesourcery.com>
22050         PR tree-optimization/83126
22051         * tree-parloops.c (num_phis): New function.
22052         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
22054 2018-03-21  Nathan Sidwell  <nathan@acm.org>
22056         * doc/extend.texi (Deprecated Features): Update deprecated flags,
22057         mention anon-struct/union members and trailing attributes.
22059 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
22061         PR tree-optimization/84969
22062         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
22063         builtin memset partitions if they set different rhs values.
22065 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
22067         PR rtl-optimization/84989
22068         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
22069         VEC_DUPLICATE with scalar result mode.
22071 2018-03-21  Martin Liska  <mliska@suse.cz>
22073         PR ipa/84963
22074         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
22075         not intended return statement.
22077 2018-03-21  Martin Liska  <mliska@suse.cz>
22079         PR target/84988
22080         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
22081         (chkp_find_bound_slots_1): Limit number of iterations.
22083 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
22085         PR target/84838
22086         * Minor grammar fixes for x86 options.
22088 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
22090         PR debug/84875
22091         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
22092         holding REG_CFA_RESTORE notes, instead turn them into a USE.
22094 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
22096         PR target/83789
22097         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
22098         (altivec_lvx_<mode>_1op): Likewise.
22099         (altivec_stvx_<mode>_2op): Likewise.
22100         (altivec_stvx_<mode>_1op): Likewise.
22101         (altivec_lvx_<VM2:mode>): New define_expand.
22102         (altivec_stvx_<VM2:mode>): Likewise.
22103         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
22104         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
22105         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
22106         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
22107         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
22108         (rs6000_gen_lvx): Likewise.
22109         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
22110         (altivec_expand_stv_builtin): Likewise.
22111         (altivec_expand_builtin): Likewise.
22112         * config/rs6000/vector.md: Likewise.
22114 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22116         PR target/82518
22117         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
22118         BYTES_BIG_ENDIAN.
22120 2018-03-20  Richard Biener  <rguenther@suse.de>
22122         PR target/84986
22123         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
22124         sign-conversions as zero, fall back to standard scalar_stmt
22125         cost for the rest.
22127 2018-03-20  Martin Liska  <mliska@suse.cz>
22129         PR ipa/84825
22130         * predict.c (rebuild_frequencies): Handle case when we have
22131         PROFILE_ABSENT, but flag_guess_branch_prob is false.
22133 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
22135         PR target/84990
22136         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
22137         flag_section_anchors.
22138         * varasm.c (use_blocks_for_decl_p): Remove hack for
22139         dw2_force_const_mem.
22141         PR target/84845
22142         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
22143         to ...
22144         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
22145         be created, use lowpart_subreg of operands[0] rather than operands[0]
22146         itself.
22147         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
22148         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
22149         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
22150         and n constraint instead of aarch64_shift_imm_di and Usd.
22151         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
22152         (*aarch64_<optab>_reg_minus<mode>3): ... this.
22154 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
22156         PR target/82989
22157         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
22158         to favor GPR over NEON registers.
22159         (<shift>di3_neon): Likewise.
22161 2018-03-20  Tom de Vries  <tom@codesourcery.com>
22163         PR target/84952
22164         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
22165         (nvptx_process_pars): Emit bar.sync asap and alap.
22167 2018-03-20  Tom de Vries  <tom@codesourcery.com>
22169         PR target/84954
22170         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
22171         seen_label if seen_label is already set.
22173 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
22175         PR target/84945
22176         * config/i386/i386.c (fold_builtin_cpu): For features above 31
22177         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
22178         Use 1U instead of 1.  Formatting fixes.
22180         PR c/84953
22181         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
22182         instead of TREE_TYPE (s1) for the return value.
22184 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
22186         PR tree-optimization/84946
22187         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
22188         bitsize + bitsize in poly_uint64 rather than poly_int64.
22190         PR sanitizer/78651
22191         * dwarf2asm.c: Include fold-const.c.
22192         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
22193         of decl rather than decl itself.
22195         PR rtl-optimization/84643
22196         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
22198 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
22200         PR sanitizer/78651
22201         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
22202         calling assemble_variable.
22204 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
22206         PR target/81647
22207         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
22208         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
22210 2018-03-19  Jim Wilson  <jimw@sifive.com>
22212         PR bootstrap/84856
22213         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
22214         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
22215         (riscv_first_stack_step): Likewise.
22216         (riscv_option_override): Use STACK_BOUNDARY instead of
22217         MIN_STACK_BOUNDARY.
22218         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
22219         MIN_STACK_BOUNDARY.
22220         (BIGGEST_ALIGNMENT): Set to 128.
22221         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
22222         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
22223         STACK_BOUNDARY.
22225 2018-03-19  Richard Biener  <rguenther@suse.de>
22227         PR tree-optimization/84933
22228         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
22229         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
22231 2018-03-19  Richard Biener  <rguenther@suse.de>
22233         PR tree-optimization/84859
22234         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
22235         (cond_if_else_store_replacement): Perform sinking operation on
22236         single-store BBs regardless of MAX_STORES_TO_SINK setting.
22237         Generalize what a BB with a single eligible store is.
22239 2018-03-19  Richard Biener  <rguenther@suse.de>
22241         PR tree-optimization/84929
22242         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
22243         chrec_is_positive against non-chrec arg.
22245 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
22247         PR target/84711
22248         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
22250 2018-03-18  Martin Liska  <mliska@suse.cz>
22252         PR rtl-optimization/84635
22253         * regrename.c (build_def_use): Use matches_mode only when
22254         matches >= 0.
22256 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
22258         PR tree-optimization/84913
22259         * tree-vect-loop.c (vectorizable_reduction): Don't try to
22260         vectorize chains of COND_EXPRs.
22262 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
22264         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
22266 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
22268         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
22270 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
22272         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
22274 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
22275             Kito Cheng  <kito.cheng@gmail.com>
22277         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
22278         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
22279         (nds32_adjust_reg_alloc_order): New function.
22280         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
22282 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
22284         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
22285         nds32_print_operand, nds32_print_operand_address): Use
22286         HOST_WIDE_INT_PRINT_DEC instead.
22288 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
22290         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
22292 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
22294         PR target/84902
22295         * config/i386/i386.c (initial_ix86_tune_features,
22296         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
22297         unsigned long long.
22298         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
22299         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
22300         rather than 1u << ix86_tune.  Formatting fix.
22301         (ix86_option_override_internal): Change ix86_arch_mask from
22302         unsigned int to unsigned HOST_WIDE_INT, initialize to
22303         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
22304         (ix86_function_specific_restore): Likewise.
22306 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
22308         PR target/84899
22309         * postreload.c (reload_combine_recognize_pattern): Perform
22310         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
22311         truncate_int_for_mode the result for the destination's mode.
22313         PR c/84909
22314         * hsa-gen.c (mem_type_for_type): Fix comment typo.
22315         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
22316         Likewise.
22317         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
22318         Likewise.
22320 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
22322         PR target/84876
22323         * lra-assigns.c (lra_split_hard_reg_for): Don't use
22324         regno_allocno_class_array and sorted_pseudos.
22325         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
22326         insns where regno is used.
22328 2018-03-16  Martin Liska  <mliska@suse.cz>
22330         PR ipa/84833
22331         * multiple_target.c (create_dispatcher_calls): Redirect
22332         reference in the symbol table.
22334 2018-03-16  Martin Liska  <mliska@suse.cz>
22336         PR ipa/84722
22337         * multiple_target.c (create_dispatcher_calls): Redirect also
22338         an alias.
22340 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
22342         PR c++/79937
22343         PR c++/82410
22344         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
22345         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
22346         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
22348 2018-03-16  Julia Koval  <julia.koval@intel.com>
22350         * doc/invoke.texi (Skylake Server): Add CLWB.
22351         Cannonlake): Remove CLWB.
22353 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
22355         PR tree-optimization/84841
22356         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
22357         1 << 3.
22358         (FLOAT_ONE_CONST_TYPE): Define.
22359         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
22360         (sort_by_operand_rank): Put entries with higher constant_type last
22361         rather than first to match comments.
22363 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
22365         * config/nios2/nios2.md (movsi_internal): Fix thinko in
22366         split predicate.
22368 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
22370         PR c++/79085
22371         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
22372         check and use address of target always.
22374 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
22376         PR target/84574
22377         * config/i386/i386.c (indirect_thunk_needed): Update comments.
22378         (indirect_thunk_bnd_needed): Likewise.
22379         (indirect_thunks_used): Likewise.
22380         (indirect_thunks_bnd_used): Likewise.
22381         (indirect_return_needed): New.
22382         (indirect_return_bnd_needed): Likewise.
22383         (output_indirect_thunk_function): Add a bool argument for
22384         function return.
22385         (output_indirect_thunk_function): Don't generate alias for
22386         function return thunk.
22387         (ix86_code_end): Call output_indirect_thunk_function to generate
22388         function return thunks.
22389         (ix86_output_function_return): Set indirect_return_bnd_needed
22390         and indirect_return_needed instead of indirect_thunk_bnd_needed
22391         and indirect_thunk_needed.
22393 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
22395         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
22396         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
22397         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
22399 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
22400             Paul Hua <paul.hua.gm@gmail.com>
22402         PR c/84852
22403         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
22405 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
22407         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
22408         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
22409         resp. SFmode cases.
22411 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
22413         PR target/84711
22414         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
22415         instead of GET_MODE_SIZE when comparing Units.
22417 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
22419         PR target/68256
22420         * varasm.c (hash_section): Return an unchangeble hash value
22421         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
22422         Return !aarch64_can_use_per_function_literal_pools_p ().
22424 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
22426         PR target/84860
22427         * optabs.c (emit_conditional_move): Pass address of cmode's copy
22428         rather than address of cmode as last argument to prepare_cmp_insn.
22430 2018-03-15  Julia Koval  <julia.koval@intel.com>
22432         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
22433         F_AVX512VNNI, F_AVX512BITALG): New.
22435 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
22437         PR target/83451
22438         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
22439         insn for floating-point loads and stores.
22441 2018-03-14  Carl Love  <cel@us.ibm.com>
22443         * config/rs6000/rs6000-c.c: Add macro definitions for
22444         ALTIVEC_BUILTIN_VEC_PERMXOR.
22445         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
22446         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
22447         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
22448         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
22449         UNSPEC_VPERMXOR.
22450         * config/doc/extend.texi: Add prototypes for vec_permxor.
22452 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
22454         PR c/84852
22455         * diagnostic-show-locus.c (class layout_point): Convert m_line
22456         from int to linenum_type.
22457         (line_span::comparator): Use linenum "compare" function when
22458         comparing line numbers.
22459         (test_line_span): New function.
22460         (layout_range::contains_point): Convert param "row" from int to
22461         linenum_type.
22462         (layout_range::intersects_line_p): Likewise.
22463         (layout::will_show_line_p): Likewise.
22464         (layout::print_source_line): Likewise.
22465         (layout::should_print_annotation_line_p): Likewise.
22466         (layout::print_annotation_line): Likewise.
22467         (layout::print_leading_fixits): Likewise.
22468         (layout::annotation_line_showed_range_p): Likewise.
22469         (struct line_corrections): Likewise for field m_row.
22470         (line_corrections::line_corrections): Likewise for param "row".
22471         (layout::print_trailing_fixits): Likewise.
22472         (layout::get_state_at_point): Likewise.
22473         (layout::get_x_bound_for_row): Likewise.
22474         (layout::print_line): Likewise.
22475         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
22476         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
22477         * input.c (selftest::test_linenum_comparisons): New function.
22478         (selftest::input_c_tests): Call it.
22479         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
22480         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
22481         * selftest.h (ASSERT_GT): New macro.
22482         (ASSERT_GT_AT): New macro.
22483         (ASSERT_LT): New macro.
22484         (ASSERT_LT_AT): New macro.
22486 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
22488         PR rtl-optimization/84780
22489         * combine.c (distribute_links): Don't make a link based on pc_rtx.
22491 2018-03-14  Martin Liska  <mliska@suse.cz>
22493         * tree.c (record_node_allocation_statistics): Use
22494         get_stats_node_kind.
22495         (get_stats_node_kind): New function extracted from
22496         record_node_allocation_statistics.
22497         (free_node): Use get_stats_node_kind.
22499 2018-03-14  Richard Biener  <rguenther@suse.de>
22501         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
22502         that the value-set of ANTIC_IN doesn't grow.
22504         Revert
22505         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
22506         member.
22507         (BB_VISITED_WITH_VISITED_SUCCS): New define.
22508         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
22510 2018-03-14  Julia Koval  <julia.koval@intel.com>
22512         * config.gcc (icelake-client, icelake-server): New.
22513         (icelake): Remove.
22514         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
22515         (initial_ix86_arch_features): Ditto.
22516         (PTA_SKYLAKE): Add SGX.
22517         (PTA_ICELAKE): Remove.
22518         (PTA_ICELAKE_CLIENT): New.
22519         (PTA_ICELAKE_SERVER): New.
22520         (ix86_option_override_internal): Split up icelake on icelake client and
22521         icelake server.
22522         (get_builtin_code_for_version): Ditto.
22523         (fold_builtin_cpu): Ditto.
22524         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
22525         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
22526         * config/i386/i386.h (processor_type): Ditto.
22527         * doc/invoke.texi: Ditto.
22529 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
22531         PR sanitizer/83392
22532         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
22533         INTEGER_CST offset, add it together with bitpos / 8 and
22534         sign extend based on POINTER_SIZE.
22536         PR target/84844
22537         Revert
22538         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
22540         PR target/78090
22541         * config/i386/constraints.md (Yc): New register constraint.
22542         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
22543         Use Yc constraint for alternative 2 of operand 0.  Remove
22544         preferred_for_speed attribute.
22546 2018-03-14  Richard Biener  <rguenther@suse.de>
22548         PR tree-optimization/84830
22549         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
22550         with the old one to avoid oscillations.
22552 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
22554         PR target/83712
22555         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
22556         pseudos.
22557         (assign_by_spills): Return a flag of reload assignment failure.
22558         Do not process the reload assignment failures.  Do not spill other
22559         reload pseudos if they has the same reg class.  Update n if
22560         necessary.
22561         (lra_assign): Add a return arg.  Set up from the result of
22562         assign_by_spills call.
22563         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
22564         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
22565         usage_insns if it is not NULL.
22566         (spill_hard_reg_in_range): New function.
22567         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
22568         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
22569         function prototypes.
22570         (lra_assign): Change prototype.
22571         * lra.c (lra): Add code to deal with fails by splitting hard reg
22572         live ranges.
22574 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
22576         * config/riscv/riscv.opt (mrelax): New option.
22577         * config/riscv/riscv.c (riscv_file_start): Emit ".option
22578         "norelax" when riscv_mrelax is disabled.
22579         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
22581 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22583         PR target/84743
22584         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
22585         reassociation for int modes.
22587 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
22589         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
22590         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
22591         for big-endian.
22592         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
22593         * config/aarch64/aarch64-sve.md
22594         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
22595         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
22596         (*extend<mode><Vwide>2): Rename to...
22597         (aarch64_sve_extend<mode><Vwide>2): ...this.
22598         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
22599         renaming the old pattern to...
22600         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
22601         unsigned packs.
22602         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
22603         define_expand, renaming the old pattern to...
22604         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
22605         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
22606         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
22607         account when deciding which SVE instruction the optab should use.
22608         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
22610 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
22612         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
22613         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
22614         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
22615         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
22616         (tlsdesc_small_<mode>): Turn a define_expand and use
22617         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
22618         (tlsdesc_small_advsimd_<mode>): ...this.
22619         (tlsdesc_small_sve_<mode>): New pattern.
22621 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
22623         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
22624         (UNSPEC_UMUL_HIGHPART): New constants.
22625         (MUL_HIGHPART): New int iteraor.
22626         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
22627         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
22628         define_expand.
22629         (*<su>mul<mode>3_highpart): New define_insn.
22631 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
22633         PR lto/84805
22634         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
22635         incomplete types.
22637 2018-03-13  Martin Liska  <mliska@suse.cz>
22639         PR ipa/84658.
22640         * (sem_item_optimizer::sem_item_optimizer): Initialize new
22641         vector.
22642         (sem_item_optimizer::~sem_item_optimizer): Release it.
22643         (sem_item_optimizer::merge_classes): Register variable aliases.
22644         (sem_item_optimizer::fixup_pt_set): New function.
22645         (sem_item_optimizer::fixup_points_to_sets): Likewise.
22646         * ipa-icf.h: Declare new variables and functions.
22648 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
22650         PR middle-end/84834
22651         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
22652         integer_pow2p@2 and test integer_pow2p in condition.
22653         (A < 0 ? C : 0): Similarly for @1.
22655         PR middle-end/84831
22656         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
22657         characters starting at p contain '\0' character, don't look beyond
22658         that.
22660         PR target/84827
22661         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
22662         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
22664         PR target/84828
22665         * reg-stack.c (change_stack): Change update_end var from int to
22666         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
22667         also call set_block_for_insn on the newly added insns and rescan.
22669         PR target/84786
22670         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
22671         on the last operand.
22673         PR c++/84704
22674         * tree.c (stabilize_reference_1): Return save_expr (e) for
22675         STATEMENT_LIST even if it doesn't have side-effects.
22677 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
22679         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
22681 2018-03-12  Renlin Li  <renlin.li@arm.com>
22683         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
22684         aarch64_output_scalar_simd_mov_immediate.
22686 2018-03-12  Martin Sebor  <msebor@redhat.com>
22688         PR tree-optimization/83456
22689         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
22690         for perfectly overlapping calls to memcpy.
22691         (gimple_fold_builtin_memory_chk): Same.
22692         (gimple_fold_builtin_strcpy): Handle no-warning.
22693         (gimple_fold_builtin_stxcpy_chk): Same.
22694         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
22696 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
22698         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
22699         parameter.  Use it for SFmode.
22700         (rs6000_function_arg_advance_1): Adjust.
22701         (rs6000_function_arg): Adjust.
22702         (rs6000_gimplify_va_arg): Pass false for that new parameter.
22704 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
22706         PR rtl-optimization/84169
22707         PR rtl-optimization/84780
22708         * combine.c (can_combine_p): Check for a 2-insn combination whether
22709         the destination register is used between the two insns, too.
22711 2018-03-12  Richard Biener  <rguenther@suse.de>
22713         PR tree-optimization/84803
22714         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
22715         for refs DR analysis didn't process.
22717 2018-03-12  Richard Biener  <rguenther@suse.de>
22719         PR tree-optimization/84777
22720         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
22721         force-vectorize loops ignore whether we are optimizing for size.
22723 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
22725         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
22726         (TARGET_MD_ASM_ADJUST): Define.
22728 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
22729             Kito Cheng  <kito.cheng@gmail.com>
22730             Chung-Ju Wu  <jasonwucj@gmail.com>
22732         * config/nds32/nds32.c (nds32_compute_stack_frame,
22733         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
22734         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
22735         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
22736         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
22737         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
22738         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
22739         * config/nds32/nds32.md (prologue, epilogue): Use macro
22740         NDS32_V3PUSH_AVAILABLE_P to do checking.
22742 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
22744         PR debug/58150
22745         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
22746         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
22747         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
22748         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
22749         addition of most attributes on !orig_type_die or the attribute not
22750         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
22752 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
22753             Chung-Ju Wu  <jasonwucj@gmail.com>
22755         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
22756         __NDS32_VH__ macro.
22757         * config/nds32/nds32.opt (mvh): New option.
22759 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
22760             Chung-Ju Wu  <jasonwucj@gmail.com>
22762         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
22763         function.
22764         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
22765         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
22766         definition.
22768 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
22769             Chung-Ju Wu  <jasonwucj@gmail.com>
22771         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
22772         function.
22773         * config/nds32/nds32-multiple.md (strlensi): New pattern.
22774         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
22776 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
22777             Kito Cheng  <kito.cheng@gmail.com>
22778             Chung-Ju Wu  <jasonwucj@gmail.com>
22780         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
22781         UNSPEC_FFMISM and UNSPEC_FLMISM.
22782         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
22783         for ffb, ffmism and flmism.
22784         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
22785         (unspec_ffmism): Ditto.
22786         (unspec_flmism): Ditto.
22787         (nds32_expand_builtin_impl): Check if string extension is available.
22788         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
22789         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
22791 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
22793         Reverting patch:
22794         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
22796         PR target/83712
22797         * lra-assigns.c (assign_by_spills): Return a flag of reload
22798         assignment failure.  Do not process the reload assignment
22799         failures.  Do not spill other reload pseudos if they has the same
22800         reg class.
22801         (lra_assign): Add a return arg.  Set up from the result of
22802         assign_by_spills call.
22803         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
22804         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
22805         usage_insns if it is not NULL.
22806         (spill_hard_reg_in_range): New function.
22807         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
22808         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
22809         function prototypes.
22810         (lra_assign): Change prototype.
22811         * lra.c (lra): Add code to deal with fails by splitting hard reg
22812         live ranges.
22814 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
22816         PR target/84807
22817         * config/i386/i386.opt: Replace Enforcment with Enforcement.
22819 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
22821         PR debug/84620
22822         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
22823         (dw_val_node): Add val_symbolic_view.
22824         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
22825         (symview_upper_bound): New.
22826         (new_line_info_table): Initialize symviews_since_reset.
22827         (dwarf2out_source_line): Count symviews_since_reset and set
22828         symview_upper_bound.
22829         (dw_val_equal_p): Handle symview.
22830         (add_AT_symview): New.
22831         (print_dw_val): Handle symview.
22832         (attr_checksum, attr_checksum_ordered): Likewise.
22833         (same_dw_val_p, size_of_die): Likewise.
22834         (value_format, output_die): Likewise.
22835         (add_high_low_attributes): Use add_AT_symview for entry_view.
22836         (dwarf2out_finish): Reset symview_upper_bound, clear
22837         zero_view_p.
22839 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
22841         PR target/83969
22842         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
22843         Add strict argument and use it.
22844         (rs6000_split_multireg_move): Update for new strict argument.
22845         (mem_operand_gpr): Disallow all non-offsettable addresses.
22846         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
22848 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
22850         PR target/84772
22851         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
22852         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
22853         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
22855         PR c++/84767
22856         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
22857         decl, use remap_type if we want to use the type.
22859 2018-03-09  Martin Sebor  <msebor@redhat.com>
22861         PR tree-optimization/84526
22862         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
22863         Remove dead code.
22864         (builtin_access::generic_overlap): Be prepared to handle non-array
22865         base objects.
22867 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
22869         PR rtl-optimization/84682
22870         * lra-constraints.c (process_address_1): Check is_address flag
22871         for address constraints.
22872         (process_alt_operands): Likewise.
22873         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
22874         preprocess_constraints.
22875         * recog.h (preprocess_constraints): Add oploc parameter.
22876         Adjust callers.
22877         * recog.c (preprocess_constraints): Test address_operand for
22878         CT_ADDRESS constraints.
22880 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
22882         PR target/83712
22883         * lra-assigns.c (assign_by_spills): Return a flag of reload
22884         assignment failure.  Do not process the reload assignment
22885         failures.  Do not spill other reload pseudos if they has the same
22886         reg class.
22887         (lra_assign): Add a return arg.  Set up from the result of
22888         assign_by_spills call.
22889         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
22890         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
22891         usage_insns if it is not NULL.
22892         (spill_hard_reg_in_range): New function.
22893         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
22894         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
22895         function prototypes.
22896         (lra_assign): Change prototype.
22897         * lra.c (lra): Add code to deal with fails by splitting hard reg
22898         live ranges.
22900 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22902         PR target/83193
22903         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
22904         Accept complain bool parameter.  Only emit errors if it is true.
22905         (arm_parse_cpu_option_name): Likewise.
22906         (arm_target_thumb_only): Adjust callers of the above.
22907         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
22908         prototype to take a default true bool parameter.
22909         (arm_parse_arch_option_name): Likewise.
22911 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
22912             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
22914         PR jit/64089
22915         PR jit/84288
22916         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
22917         * configure: Regenerate.
22918         * configure.ac ("linker --version-script option"): New.
22919         ("linker soname option"): New.
22921 2018-03-09  Richard Biener  <rguenther@suse.de>
22923         PR tree-optimization/84775
22924         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
22925         immediate uses of predicate stmts and mark them modified.
22927         Revert
22928         PR tree-optimization/84178
22929         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
22930         to caller.
22931         (version_loop_for_if_conversion): Delay update_ssa call.
22932         (tree_if_conversion): Delay update_ssa until after predicate
22933         insertion.
22935 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
22937         PR target/84763
22938         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
22939         when the function accesses prior frames.
22941 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
22943         PR debug/84456
22944         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
22945         gen_llsym, otherwise call maybe_gen_llsym.
22947         PR inline-asm/84742
22948         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
22949         has ',' character inside of it.
22951 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22953         PR target/84748
22954         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
22955         as clobbering CC_REGNUM.
22957 2018-03-08  Richard Biener  <rguenther@suse.de>
22959         PR middle-end/84552
22960         * tree-scalar-evolution.c: Include tree-into-ssa.h.
22961         (follow_copies_to_constant): Do not follow SSA names registered
22962         for update.
22964 2018-03-08  Richard Biener  <rguenther@suse.de>
22966         PR tree-optimization/84178
22967         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
22968         to caller.
22969         (version_loop_for_if_conversion): Delay update_ssa call.
22970         (tree_if_conversion): Delay update_ssa until after predicate
22971         insertion.
22973 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
22975         PR tree-optimization/84178
22976         * tree-if-conv.c (release_bb_predicate): Remove the
22977         the assertion that the stmts have NULL use_ops.
22978         Discard the statements, asserting that they haven't
22979         yet been added to a BB.
22981 2018-03-08  Richard Biener  <rguenther@suse.de>
22983         PR tree-optimization/84746
22984         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
22985         (phi_translate): Pass in destination ANTIC_OUT set.
22986         (phi_translate_1): Likewise.  For a simplified result lookup
22987         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
22988         (phi_translate_set): Adjust.
22989         (do_pre_regular_insertion): Likewise.
22990         (do_pre_partial_partial_insertion): Likewise.
22992 2018-03-08  Martin Liska  <mliska@suse.cz>
22994         PR gcov-profile/84735
22995         * doc/gcov.texi: Document usage of profile files.
22996         * gcov-io.h: Document changes in the format.
22998 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
23000         PR debug/84404
23001         PR debug/84408
23002         * dwarf2out.c (struct dw_line_info_table): Update comments for
23003         view == -1.
23004         (FORCE_RESET_NEXT_VIEW): New.
23005         (FORCE_RESETTING_VIEW_P): New.
23006         (RESETTING_VIEW_P): Check for -1 too.
23007         (ZERO_VIEW_P): Likewise.
23008         (new_line_info_table): Force-reset next view.
23009         (dwarf2out_begin_function): Likewise.
23010         (dwarf2out_source_line): Simplify zero_view_p initialization.
23011         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
23012         view directly.  Omit view when omitting .loc at line 0.
23014 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
23016         PR tree-optimization/84740
23017         * tree-switch-conversion.c (process_switch): Call build_constructors
23018         only if info.phi_count is non-zero.
23020         PR tree-optimization/84739
23021         * tree-tailcall.c (find_tail_calls): Check call arguments against
23022         DECL_ARGUMENTS (current_function_decl) rather than
23023         DECL_ARGUMENTS (func) when checking for tail recursion.
23025 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
23027         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
23028         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
23029         Volker Reichelt's entry and add entries for people that perform
23030         GCC fuzzy testing and report numerous bugs.
23032 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
23034         PR target/82411
23035         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
23036         readonly data in sdata, if that is disabled.
23037         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
23038         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
23039         -mreadonly-in-sdata option.
23041 2018-03-07  Martin Sebor  <msebor@redhat.com>
23043         PR tree-optimization/84468
23044         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
23045         basic block when looking for nul assignment.
23047 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
23049         PR target/84277
23050         * except.h (output_function_exception_table): Adjust prototype.
23051         * except.c (output_function_exception_table): Remove FNNAME parameter
23052         and add SECTION parameter.  Ouput one part of the table at a time.
23053         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
23054         the first part of the exception table and emit unwind directives.
23055         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
23056         (i386_pe_seh_cold_init): Likewise.
23057         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
23058         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
23059         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
23060         (ix86_output_call_insn): Emit a nop in one more case for SEH.
23061         * config/i386/winnt.c: Include except.h.
23062         (struct seh_frame_state): Add reg_offset, after_prologue and
23063         in_cold_section fields.
23064         (i386_pe_seh_end_prologue): Set seh->after_prologue.
23065         (i386_pe_seh_cold_init): New function.
23066         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
23067         to seh->in_cold_section.
23068         (seh_emit_push): Record the offset of the push.
23069         (seh_emit_save): Record the offet of the save.
23070         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
23071         Test seh->after_prologue to disregard the epilogue.
23072         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
23073         (i386_pe_end_cold_function): New function.
23075 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
23077         PR fortran/84565
23078         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
23079         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
23081         PR c++/84704
23082         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
23083         on tmp_var.
23084         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
23085         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
23087         PR middle-end/84723
23088         * multiple_target.c: Include tree-inline.h and intl.h.
23089         (expand_target_clones): Diagnose and fail if node->definition and
23090         !tree_versionable_function_p (node->decl).
23092 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
23094         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
23095         sprint_ul.
23096         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
23097         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
23098         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
23100 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
23102         PR target/84710
23103         * combine.c (try_combine): Use reg_or_subregno instead of handling
23104         just paradoxical SUBREGs and REGs.
23106 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
23108         * config/arc/arc.c (arc_finalize_pic): Remove function.
23109         (arc_must_save_register): We use single base PIC register, remove
23110         checks to save/restore the PIC register.
23111         (arc_expand_prologue): Likewise.
23112         * config/arc/arc-protos.h (arc_set_default_type_attributes):
23113         Remove.
23114         (arc_verify_short): Likewise.
23115         (arc_attr_type): Likewise.
23116         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
23117         (walk_stores): Likewise.
23118         (arc_address_cost): Make it static.
23119         (arc_verify_short): Likewise.
23120         (branch_dest): Likewise.
23121         (arc_attr_type): Likewise.
23122         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
23123         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
23124         (arc_final_prescan_insn): Remove inserting the nops due to
23125         hardware hazards.  It is done in reorg step.
23126         (insn_length_variant_t): Remove.
23127         (insn_length_parameters_t): Likewise.
23128         (arc_insn_length_parameters): Likewise.
23129         (arc_get_insn_variants): Likewise.
23130         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
23132 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
23134         PR inline-asm/84683
23135         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
23136         assertion failure.
23138         PR tree-optimization/84687
23139         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
23140         on new_node->decl.
23141         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
23143 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23145         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
23146         Rename to ppc_speculation_barrier.
23147         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
23148         __builtin_ppc_speculation_barrier.
23150 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
23152         PR target/84700
23153         * combine.c (combine_simplify_rtx): Don't try to simplify if
23154         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
23155         are equal to x.
23157 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
23159         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
23160         to 32 bytes when compiling for POWER9.
23162 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
23164         PR target/84564
23165         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
23166         regparm >= 3 with no arg reg available also for calls with
23167         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
23169         PR target/84524
23170         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
23171         orig,vex.
23172         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
23174 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
23176         PR target/84264
23177         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
23179 2018-03-05  Richard Biener  <rguenther@suse.de>
23181         PR tree-optimization/84486
23182         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
23183         When inserting a __builtin_assume_aligned call set the LHS
23184         SSA name alignment info accordingly.
23186 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
23188         PR tree-optimization/84114
23189         * config/aarch64/aarch64.c (aarch64_reassociation_width)
23190         Avoid reassociation of FLOAT_MODE addition.
23192 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
23194         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
23195         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
23196         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
23197         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
23198         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
23199         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
23200         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
23201         and -mwbnoinvd.
23202         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
23203         __builtin_ia32_wbinvd): New builtins.
23204         (SPECIAL_ARGS2): New.
23205         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
23206         (SPECIAL_ARGS2): New.
23207         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
23208         (ix86_valid_target_attribute_inner_p): Ditto.
23209         (ix86_init_mmx_sse_builtins): Add special_args2.
23210         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
23211         TARGET_WBNOINVD_P): New.
23212         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
23213         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
23214         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
23215         * config/i386/immintrin.h (_wbinvd): New intrinsic.
23216         * config/i386/pconfigintrin.h: New file.
23217         * config/i386/wbnoinvdintrin.h: Ditto.
23218         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
23219         wbnoinvdintrin.h.
23220         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
23222 2018-03-05  Richard Biener  <rguenther@suse.de>
23224         PR tree-optimization/84670
23225         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
23226         member.
23227         (BB_VISITED_WITH_VISITED_SUCCS): New define.
23228         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
23229         (compute_antic_aux): Only assert the number of values in ANTIC_IN
23230         doesn't grow if all successors (recursively) were visited at least
23231         once.
23233 2018-03-05  Richard Biener  <rguenther@suse.de>
23235         PR tree-optimization/84650
23236         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
23237         if executed in the loop pipeline.
23239 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
23241         * doc/configfiles.texi (Configuration Files): Move info about
23242         conditionalizing $target-protos.h to...
23243         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
23244         differs from $target-protos.h.
23246 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
23247             Chung-Ju Wu  <jasonwucj@gmail.com>
23249         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
23250         * config/nds32/nds32-multiple.md (setmemsi): Define.
23251         * config/nds32/nds32-memory-manipulation.c
23252         (nds32_gen_dup_4_byte_to_word_value): New.
23253         (emit_setmem_word_loop): New.
23254         (emit_setmem_byte_loop): New.
23255         (nds32_expand_setmem_loop): New.
23256         (nds32_expand_setmem_loop_v3m): New.
23257         (nds32_expand_setmem_unroll): New.
23258         (nds32_expand_setmem): New.
23260 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
23261             Chung-Ju Wu  <jasonwucj@gmail.com>
23263         * config/nds32/nds32-memory-manipulation.c
23264         (nds32_emit_load_store): New.
23265         (nds32_emit_post_inc_load_store): New.
23266         (nds32_emit_mem_move): New.
23267         (nds32_emit_mem_move_block): New.
23268         (nds32_expand_movmemsi_loop_unknown_size): New.
23269         (nds32_expand_movmemsi_loop_known_size): New.
23270         (nds32_expand_movmemsi_loop): New.
23271         (nds32_expand_movmemsi_unroll): New.
23272         (nds32_expand_movmemqi): Rename ...
23273         (nds32_expand_movmemsi): ... to this.
23274         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
23275         (movmemsi): ... to this.
23276         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
23277         (nds32_expand_movmemsi): ... to this.
23279 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
23280             Monk Chiang  <sh.chiang04@gmail.com>
23281             Chung-Ju Wu  <jasonwucj@gmail.com>
23283         * config/nds32/nds32-protos.h
23284         (nds32_expand_load_multiple): New arguments.
23285         (nds32_expand_store_multiple): Ditto.
23286         (nds32_valid_multiple_load_store): Rename ...
23287         (nds32_valid_multiple_load_store_p): ... to this.
23288         * config/nds32/nds32-memory-manipulation.c
23289         (nds32_expand_load_multiple): Refine implementation.
23290         (nds32_expand_store_multiple): Ditto.
23291         * config/nds32/nds32-multiple.md
23292         (load_multiple): Update nds32_expand_load_multiple interface.
23293         (store_multiple): Update nds32_expand_store_multiple interface.
23294         * config/nds32/nds32-predicates.c
23295         (nds32_valid_multiple_load_store): Rename ...
23296         (nds32_valid_multiple_load_store_p): ... to this and refine
23297         implementation.
23298         * config/nds32/predicates.md
23299         (nds32_load_multiple_and_update_address_operation): New predicate.
23300         (nds32_store_multiple_and_update_address_operation): New predicate.
23302 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
23303             Chung-Ju Wu  <jasonwucj@gmail.com>
23305         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
23306         (combo): New attribute.
23307         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
23309 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
23311         * config/nds32/nds32.opt: Change -mcmodel= default value.
23313 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
23314             Monk Chiang  <sh.chiang04@gmail.com>
23315             Chung-Ju Wu  <jasonwucj@gmail.com>
23317         * config/nds32/constants.md (unspec_element): New enum.
23318         * config/nds32/constraints.md (Umw): New constraint.
23319         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
23320         * config/nds32/nds32-intrinsic.md: Likewise.
23321         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
23322         (nds32_valid_smw_lwm_base_p): New.
23323         (nds32_output_smw_single_word): New.
23324         (nds32_output_lmw_single_word): New.
23325         (nds32_expand_unaligned_load): New.
23326         (nds32_expand_unaligned_store): New.
23327         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
23328         (nds32_output_smw_single_word): Declare.
23329         (nds32_output_lmw_single_word): Declare.
23330         (nds32_expand_unaligned_load): Declare.
23331         (nds32_expand_unaligned_store): Declare.
23332         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
23333         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
23334         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
23335         NDS32_BUILTIN_UASTORE_DW.
23336         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
23337         predicate.
23339 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
23340             Kito Cheng  <kito.cheng@gmail.com>
23341             Chung-Ju Wu  <jasonwucj@gmail.com>
23343         * config/nds32/nds32-intrinsic.c
23344         (nds32_expand_builtin_null_ftype_reg): Delete.
23345         (nds32_expand_builtin_reg_ftype_imm): Ditto.
23346         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
23347         (nds32_read_argument): New.
23348         (nds32_legitimize_target): Ditto.
23349         (nds32_legitimize_argument): Ditto.
23350         (nds32_check_constant_argument): Ditto.
23351         (nds32_expand_unop_builtin): Ditto.
23352         (nds32_expand_unopimm_builtin): Ditto.
23353         (nds32_expand_binop_builtin): Ditto.
23354         (nds32_builtin_decl_impl): Ditto.
23355         (builtin_description): Ditto.
23356         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
23357         (nds32_init_builtins_impl): Ditto.
23358         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
23359         (nds32_builtin_decl): New.
23360         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
23361         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
23363 2018-03-02  Jeff Law  <law@redhat.com>
23365         * reorg.c (stop_search_p): Handle DEBUG_INSN.
23366         (redundant_insn, fill_simple_delay_slots): Likewise.
23367         (fill_slots_from_thread): Likewise.
23368         * resource.c (mark_referenced_resources): Likewise.
23369         (mark_set_resources, find_dead_or_set_registers): Likewise.
23371 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
23373         * substring-locations.h (format_warning_va): Formatting fix for
23374         ATTRIBUTE_GCC_DIAG.
23375         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
23376         argument.
23377         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
23378         * substring-locations.c: Include intl.h.
23379         (format_warning_va): Turned into small wrapper around
23380         format_warning_n_va, renamed to ...
23381         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
23382         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
23383         use ngettext.
23384         (format_warning_at_substring_n): New function.
23385         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
23386         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
23387         format_warning_at_substring with just a shorter name instead of
23388         const function pointer.
23389         (fmtwarn_n): New function.
23390         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
23391         appropriate, get rid of all the fmtstr temporaries, move conditionals
23392         with G_() wrapped string literals directly into fmtwarn arguments,
23393         cast dir.len to (int), formatting fixes.
23395 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
23397         * doc/invoke.texi: Remove "Cilk Plus" references.
23399 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
23400             Richard Biener  <rguenther@suse.de>
23402         PR ipa/84628
23403         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
23404         for error or warning attributes if CALL_FROM_THUNK_P is set.
23405         Formatting fixes.
23407 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
23409         PR target/56540
23410         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
23411         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
23413         PR target/56540
23414         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
23415         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
23417         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
23418         instead of -1U in last predictors element's probability member.
23420 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
23422         PR ipa/83983
23423         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
23424         arguments if they are comparable.
23426 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
23428         PR tree-optimization/84634
23429         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
23430         masks and masked_loop_p with a single loop_masks, making sure it's
23431         null for bb vectorization.
23433 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
23435         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
23436         (vect_analyze_data_ref_access): Use loop->safe_len rather than
23437         loop->force_vectorize to check whether there is no alias.
23439 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
23441         PR target/84614
23442         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
23443         prototypes.
23444         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
23445         comments.
23446         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
23447         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
23448         instead of a loop around prev_real_insn.
23449         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
23450         prev_real_insn.
23452         PR inline-asm/84625
23453         * config/i386/i386.c (ix86_print_operand): Use conditional
23454         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
23455         zero vector.
23457 2018-03-02  Richard Biener  <rguenther@suse.de>
23459         PR tree-optimization/84427
23460         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
23461         (bitmap_set_subtract_values): Rewrite to handle multiple
23462         exprs per value.
23463         (clean): Likewise.
23464         (prune_clobbered_mems): Likewise.
23465         (phi_translate): Take edge instead of pred/phiblock.
23466         (phi_translate_1): Likewise.
23467         (phi_translate_set): Likewise.  Insert all translated
23468         exprs for a value into the set, keeping possibly multiple
23469         expressions per value.
23470         (compute_antic_aux): Adjust for phi_translate changes.
23471         When intersecting union the expressions and prune those
23472         not in the final value set, keeping possibly multiple
23473         expressions per value.  Do not use value-insertion
23474         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
23475         all expressions.  Add verification that the value-sets
23476         only shrink during iteration.
23477         (compute_partial_antic_aux): Adjust for the phi_translate changes.
23478         (do_pre_regular_insertion): Likewise.
23479         (do_pre_partial_partial_insertion): Likewise.
23481 2018-03-02  Richard Biener  <rguenther@suse.de>
23483         PR target/82005
23484         * config/darwin.c (saved_debug_info_level): New static global.
23485         (darwin_asm_lto_start): Disable debug info generation for LTO out.
23486         (darwin_asm_lto_end): Restore debug info generation settings.
23488 2018-03-01  Martin Liska  <mliska@suse.cz>
23490         PR sanitizer/82484
23491         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
23492         volatile arguments.
23494 2018-03-01  Richard Biener  <rguenther@suse.de>
23496         PR debug/84645
23497         * dwarf2out.c (gen_variable_die): Properly handle late VLA
23498         type annotation with LTO when debug was disabled at compile-time.
23500 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
23502         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
23503         XINT with INTVAL.
23504         (mips_final_postscan_insn): Likewise.
23506 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
23508         PR rtl-optimization/84528
23509         * alias.c (init_alias_target): Add commentary.
23510         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
23511         a unique base value if the frame pointer is not eliminated
23512         to the stack pointer.
23514 2018-03-01  Tom de Vries  <tom@codesourcery.com>
23516         PR rtl-optimization/83327
23517         * lra-int.h (hard_regs_spilled_into): Declare.
23518         * lra.c (hard_regs_spilled_into): Define.
23519         (init_reg_info): Init hard_regs_spilled_into.
23520         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
23521         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
23522         (process_bb_lives): Handle hard_regs_spilled_into.
23523         (lra_create_live_ranges_1): Before doing liveness propagation, clear
23524         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
23526 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
23528         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
23529         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
23530         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
23531         * config/rs6000/aix72.h: New file.
23533 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
23535         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
23536         instead of warning_at with conditional singular and plural messages
23537         where possible.
23539         PR target/52991
23540         * stor-layout.c (update_alignment_for_field): For
23541         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
23542         && !DECL_PACKED (field), do the alignment update, just use
23543         only desired_align instead of MAX (type_align, desired_align)
23544         as the alignment.
23545         (place_field): Don't do known_align < desired_align handling
23546         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
23547         is non-NULL, instead do it after rli->prev_field handling and
23548         only if not within a bitfield word.  For DECL_PACKED (field)
23549         use type_align of BITS_PER_UNIT.
23551 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
23553         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
23554         superfluous parentheses and trailing spaces.
23556 2018-02-28  Richard Biener  <rguenther@suse.de>
23558         PR tree-optimization/84584
23559         * graphite-scop-detection.c (scop_detection::add_scop): Discard
23560         SCoPs with fake exit edge.
23562 2018-02-28  Martin Liska  <mliska@suse.cz>
23564         PR testsuite/84597
23565         * timevar.c (timer::print): Fix format to properly print 100%
23566         values.
23568 2018-02-28  Richard Biener  <rguenther@suse.de>
23570         PR middle-end/84607
23571         * genmatch.c (capture_info::walk_match): Do not mark
23572         captured expressions without operands as expr_p given
23573         they act more like predicates and should be subject to
23574         "lost tail" side-effect preserving.
23576 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
23578         PR rtl-optimization/81611
23579         * auto-inc-dec.c (attempt_change): Move dead note from
23580         mem_insn if it's the next use of regno
23581         (find_address): Take address use of reg holding
23582         non-incremented value.  Add parm to limit search to the named
23583         reg only.
23584         (merge_in_block): Attempt to use a mem insn that is the next
23585         use of the original regno.
23587 2018-02-27  Martin Sebor  <msebor@redhat.com>
23589         PR c++/83871
23590         * doc/invoke.texi (-Wmissing-attributes): New option.
23591         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
23593 2018-02-27  Martin Sebor  <msebor@redhat.com>
23595         PR translation/84207
23596         * diagnostic-core.h (warning_n, error_n, inform_n): Change
23597         n argument to unsigned HOST_WIDE_INT.
23598         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
23599         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
23600         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
23601         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
23603 2018-02-27  Richard Biener  <rguenther@suse.de>
23605         PR tree-optimization/84512
23606         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
23607         Do not use the estimate returned from record_stmt_cost for
23608         the scalar iteration cost but sum properly using add_stmt_cost.
23610 2018-02-27  Richard Biener  <rguenther@suse.de>
23612         PR tree-optimization/84466
23613         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
23614         Adjust last change to less strictly validate use operands.
23616 2018-02-27  Martin Liska  <mliska@suse.cz>
23618         PR gcov-profile/84548
23619         * gcov.c (process_file): Allow partial overlap and consider it
23620         also as group functions.
23621         (output_lines): Properly calculate range of lines for a group.
23623 2018-02-27  Martin Liska  <mliska@suse.cz>
23625         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
23626         'ggc' suffixes.  Change first column width.
23627         (timer::print): Fix formatting of the column.
23629 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
23631         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
23632         preserve inline entry blocks for the sake of debug inline
23633         entry point markers alone.
23634         (remove_unused_locals): Suggest in comments a better place to
23635         force the preservation of inline entry blocks that are
23636         otherwise unused, but do not preserve them.
23638 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
23640         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
23642 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
23644         PR target/84039
23645         * config/i386/constraints.md (Bs): Replace
23646         ix86_indirect_branch_register with
23647         TARGET_INDIRECT_BRANCH_REGISTER.
23648         (Bw): Likewise.
23649         * config/i386/i386.md (indirect_jump): Likewise.
23650         (tablejump): Likewise.
23651         (*sibcall_memory): Likewise.
23652         (*sibcall_value_memory): Likewise.
23653         Peepholes of indirect call and jump via memory: Likewise.
23654         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
23655         (*sibcall_value_GOT_32): Likewise.
23656         * config/i386/predicates.md (indirect_branch_operand): Likewise.
23657         (GOT_memory_operand): Likewise.
23658         (call_insn_operand): Likewise.
23659         (sibcall_insn_operand): Likewise.
23660         (GOT32_symbol_operand): Likewise.
23661         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
23663 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
23665         PR rtl-optimization/83496
23666         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
23667         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
23668         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
23669         redundant insn, if any.
23670         (relax_delay_slots): Likewise.
23671         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
23673 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
23675         PR tree-optimization/83965
23676         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
23677         that grouped statements are part of a reduction chain.  Return
23678         true if the statement is not marked as a reduction itself but
23679         is part of a group.
23680         (vect_recog_dot_prod_pattern): Don't check whether the statement
23681         is part of a group here.
23682         (vect_recog_sad_pattern): Likewise.
23683         (vect_recog_widen_sum_pattern): Likewise.
23685 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
23687         PR debug/84545
23688         * final.c (rest_of_clean_state): Also look for calls inside sequences.
23690 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
23692         PR target/84530
23693         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
23694         the bool argument.
23695         (ix86_output_indirect_function_return): New prototype.
23696         (ix86_split_simple_return_pop_internal): Likewise.
23697         * config/i386/i386.c (indirect_return_via_cx): New.
23698         (indirect_return_via_cx_bnd): Likewise.
23699         (indirect_thunk_name): Handle return va CX_REG.
23700         (output_indirect_thunk_function): Create alias for
23701         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
23702         (ix86_output_indirect_jmp): Remove the bool argument.
23703         (ix86_output_indirect_function_return): New function.
23704         (ix86_split_simple_return_pop_internal): Likewise.
23705         * config/i386/i386.md (*indirect_jump): Don't pass false
23706         to ix86_output_indirect_jmp.
23707         (*tablejump_1): Likewise.
23708         (simple_return_pop_internal): Change it to define_insn_and_split.
23709         Call ix86_split_simple_return_pop_internal to split it for
23710         -mfunction-return=.
23711         (simple_return_indirect_internal): Call
23712         ix86_output_indirect_function_return instead of
23713         ix86_output_indirect_jmp.
23715 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
23717         PR bootstrap/84405
23718         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
23719         memset and value initialization afterwards.
23721 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
23723         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
23725 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
23727         PR target/84521
23728         * common/config/aarch64/aarch64-common.c
23729         (aarch_option_optimization_table[]): Switch
23730         off fomit-frame-pointer
23732 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
23733             Chung-Ju Wu  <jasonwucj@gmail.com>
23735         * config/nds32/nds32-multiple.md (load_multiple): Disallow
23736         volatile memory.
23737         (store_multiple): Ditto.
23739 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
23741         * config.gcc: Add --with-cpu support for nds32 target.
23742         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
23743         * config/nds32/nds32.opt: Add -mcpu= option.
23745 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
23747         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
23748         isel=yes): Warn for these deprecated options.
23750 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
23752         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
23753         ISA_2_5_MASKS_EMBEDDED.
23755 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
23757         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
23758         p->max as pointers rather than using iterative_hash_expr.
23760 2018-02-23  Carl Love  <cel@us.ibm.com>
23762         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
23763         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
23764         BU_P8V_OVERLOAD_2.
23765         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
23766         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
23767         P8V_BUILTIN_VEC_VUNSIGNED2.
23769 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
23771         PR target/81572
23772         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
23773         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
23774         LRA_UNKNOWN_ALT.
23775         * lra-constraints.c (curr_insn_transform): Set up
23776         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
23777         LRA_UNKNOWN_ALT.
23778         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
23779         * lra-eliminations.c (spill_pseudos): Ditto.
23780         (process_insn_for_elimination): Ditto.
23781         * lra-lives.c (reg_early_clobber_p): Use the new macros.
23782         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
23783         LRA_NON_CLOBBERED_ALT.
23785 2018-02-22  Martin Sebor  <msebor@redhat.com>
23787         PR tree-optimization/84480
23788         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
23789         to maybe_diag_stxncpy_trunc.  Call it.
23790         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
23791         from gimple_fold_builtin_strcpy.  Print inlining stack.
23792         (handle_builtin_stxncpy): Print inlining stack.
23793         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
23795 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
23797         PR target/84176
23798         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
23799         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
23800         and -fcheck-pointer-bounds are used together.
23801         (indirect_thunk_prefix): New enum.
23802         (indirect_thunk_need_prefix): New function.
23803         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
23804         "_nt" instead of "_bnd" for NOTRACK prefix.
23805         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
23806         (output_indirect_thunk_function): Likewise.
23807         (): Likewise.
23808         (ix86_code_end): Update output_indirect_thunk_function calls.
23809         (ix86_output_indirect_branch_via_reg): Replace
23810         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
23811         (ix86_output_indirect_branch_via_push): Likewise.
23812         (ix86_output_function_return): Likewise.
23813         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
23814         incompatible with -fcf-protection=branch and
23815         -fcheck-pointer-bounds.
23817 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
23819         PR target/83335
23820         * config/aarch64/aarch64.c (aarch64_print_address_internal):
23821         Change gcc_assert call to output_operand_lossage.
23823 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
23825         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
23827 2018-02-22  DJ Delorie  <dj@redhat.com>
23828             Sebastian Perta  <sebastian.perta@renesas.com>
23829             Oleg Endo  <olegendo@gcc.gnu.org>
23831         * config/rx/rx.c (rx_rtx_costs): New function.
23832         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
23834 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
23836         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
23838 2018-02-22  Martin Liska  <mliska@suse.cz>
23840         PR driver/83193
23841         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
23842         Add "native" as a possible value.
23844 2018-02-22  Martin Liska  <mliska@suse.cz>
23846         PR driver/83193
23847         * config/i386/i386.c (ix86_option_override_internal):
23848         Add "native" as a possible value for -march and -mtune.
23850 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
23852         PR target/84502
23853         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
23854         to all type variants.
23856         PR tree-optimization/84503
23857         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
23858         width as info->bitpos + info->bitsize - start.
23859         (merged_store_group::merge_overlapping): Simplify width computation.
23860         (check_no_overlap): New function.
23861         (imm_store_chain_info::try_coalesce_bswap): Compute expected
23862         start + width and last_order of the group, fail if check_no_overlap
23863         fails.
23864         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
23865         to group if check_no_overlap fails.
23867 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
23869         * config/rs6000/altivec.md: Delete contraint arguments to
23870         define_expand, define_split, and define_peephole2, and in
23871         define_insn_and_split if always unused.
23872         * config/rs6000/darwin.md: Ditto.
23873         * config/rs6000/dfp.md: Ditto.
23874         * config/rs6000/rs6000.md: Ditto.
23875         * config/rs6000/sync.md: Ditto.
23876         * config/rs6000/vector.md: Ditto.
23877         * config/rs6000/vsx.md: Ditto.
23879 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
23881         * config/rs6000/altivec.md: Write output control strings as braced
23882         blocks instead of double-quoted strings.
23883         * config/rs6000/darwin.md: Ditto.
23884         * config/rs6000/rs6000.md: Ditto.
23885         * config/rs6000/vector.md: Ditto.
23886         * config/rs6000/vsx.md: Ditto.
23888 2018-02-21  Jason Merrill  <jason@redhat.com>
23890         PR c++/84314 - ICE with templates and fastcall attribute.
23891         * attribs.c (build_type_attribute_qual_variant): Remove assert.
23893 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
23895         * ipa-cp.c (determine_versionability): Fix comment typos.
23897 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
23899         PR c/84229
23900         * ipa-cp.c (determine_versionability): Do not version functions caling
23901         va_arg_pack.
23903 2018-02-21  Martin Liska  <mliska@suse.cz>
23905         PR driver/83193
23906         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
23907         Add "native" as a possible value.
23908         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
23909         the macro when native cpu detection is available.
23911 2018-02-21  Martin Liska  <mliska@suse.cz>
23913         PR driver/83193
23914         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
23915         Add "native" as a possible value.
23916         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
23917         when native cpu detection is available.
23919 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
23920             Martin Sebor  <msebor@redhat.com>
23922         PR tree-optimization/84478
23923         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
23924         false.
23925         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
23926         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
23927         support which is conservatively correct, for 2 only stay conservative
23928         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
23929         argument to the 2 argument get_range_strlen, adjust 6 arg
23930         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
23931         false.
23932         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
23933         (gimple_fold_builtin_strlen): Pass true as last argument to
23934         get_range_strlen.
23936 2018-02-20  Martin Sebor  <msebor@redhat.com>
23938         PR middle-end/84095
23939         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
23940         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
23941         (builtin_memref::builtin_memref): Factor out parts into
23942         set_base_and_offset and call it.
23944 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
23946         PR middle-end/84406
23947         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
23948         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
23949         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
23950         search at the associated MODE_INT.
23952 2018-02-20  Jeff Law  <law@redhat.com>
23954         PR middle-end/82123
23955         PR tree-optimization/81592
23956         PR middle-end/79257
23957         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
23958         for range data rather than using global data.
23959         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
23960         range data rather than using global data.
23961         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
23962         pass it to children as needed.
23963         (struct directive::fmtresult): Similarly.
23964         (struct directive::set_width): Similarly.
23965         (struct directive::set_precision): Similarly.
23966         (format_integer, format_directive, parse_directive): Similarly.
23967         (format_none): Accept unnamed vr_values parameter.
23968         (format_percent, format_floating, format_character): Similarly.
23969         (format_string, format_plain): Similarly.
23970         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
23971         the EVRP range analyzer for range data rather than using global data.
23972         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
23973         gimple-ssa-evrp-analyze.h
23974         (class sprintf_dom_walker): Add after_dom_children member function.
23975         Add evrp_range_analyzer member.
23976         (sprintf_dom_walker::before_dom_children): Call into the EVRP
23977         range analyzer as needed.
23978         (sprintf_dom_walker::after_dom_children): New member function.
23979         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
23980         if not optimizing.
23981         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
23982         (evrp_range_analyzer::pop_to_marker): Likewise.
23984 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
23986         PR tree-optimization/84419
23987         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
23988         with the required type if its current type is compatible but
23989         different.
23991 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
23993         PR middle-end/82004
23994         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
23995         after vectorization.
23997 2018-02-20  Martin Liska  <mliska@suse.cz>
23999         PR driver/83193
24000         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
24001         possible values if we don't have a hint.
24003 2018-02-20  Martin Liska  <mliska@suse.cz>
24005         PR c/84310
24006         PR target/79747
24007         * final.c (shorten_branches): Build align_tab array with one
24008         more element.
24009         * opts.c (finish_options): Add alignment option limit check.
24010         (MAX_CODE_ALIGN): Likewise.
24011         (MAX_CODE_ALIGN_VALUE): Likewise.
24012         * doc/invoke.texi: Document maximum allowed option value for
24013         all -falign-* options.
24015 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
24017         PR target/84146
24018         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
24019         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
24020         * var-tracking.c (emit_note_insn_var_location): Remove all references
24021         to NOTE_INSN_CALL_ARG_LOCATION.
24022         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
24023         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
24024         Use copy_rtx_if_shared.
24025         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
24026         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
24027         (dwarf2out_var_location): Remove handling of
24028         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
24029         on call_insn.
24030         * final.c (final_scan_insn): Remove all references to
24031         NOTE_INSN_CALL_ARG_LOCATION.
24032         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
24033         before dumping final insns.
24034         * except.c (emit_note_eh_region_end): Remove all references to
24035         NOTE_INSN_CALL_ARG_LOCATION.
24036         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
24037         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
24038         * config/arc/arc.c (hwloop_optimize): Likewise.
24039         * config/arm/arm.c (create_fix_barrier): Likewise.
24040         * config/s390/s390.c (s390_chunkify_start): Likewise.
24041         * config/sh/sh.c (find_barrier): Likewise.
24042         * config/i386/i386.c (rest_of_insert_endbranch,
24043         ix86_seh_fixup_eh_fallthru): Likewise.
24044         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
24045         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
24046         * config/frv/frv.c (frv_function_prologue): Likewise.
24047         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
24048         reg note.
24049         (note_outside_basic_block_p): Remove all references to
24050         NOTE_INSN_CALL_ARG_LOCATION.
24051         * gengtype.c (adjust_field_rtx_def): Likewise.
24052         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
24053         Likewise.
24054         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
24055         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
24057         PR c++/84444
24058         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
24059         is ADDR_EXPR.
24061         PR tree-optimization/84452
24062         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
24063         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
24064         is NULL.
24066 2018-02-19  Martin Liska  <mliska@suse.cz>
24068         PR sanitizer/82183
24069         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
24071 2018-02-19  Martin Liska  <mliska@suse.cz>
24072             Richard Sandiford  <richard.sandiford@linaro.org>
24074         PR tree-optimization/82491
24075         * gimple-fold.c (get_base_constructor): Make earlier bail out
24076         to prevent ubsan.
24078 2018-02-19  Carl Love  <cel@us.ibm.com>
24080         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
24081         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
24082         BU_P8V_OVERLOAD_1.
24083         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
24084         P8V_BUILTIN_VEC_NEG.
24086 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
24088         * config/rl78/rl78.md (movdf): New define expand.
24090 2018-02-19  Martin Liska  <mliska@suse.cz>
24092         PR other/80589
24093         * doc/invoke.texi: Fix typo.
24094         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
24096 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
24098         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
24099         handle rs6000_single_float and rs6000_double_float specially for
24100         e500 family CPUs.
24102 2018-02-16  Jeff Law  <law@redhat.com>
24104         * config/rx/rx.c (add_pop_cfi_notes): New function.;
24105         (pop_regs): Use it.
24107 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
24109         PR ipa/84425
24110         * ipa-inline.c (inline_small_functions): Fix a typo.
24112 2018-02-16  Nathan Sidwell  <nathan@acm.org>
24114         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
24116 2018-02-16  Carl Love  <cel@us.ibm.com>
24118         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
24119         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
24120         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
24121         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
24122         expansion to P8V_BUILTIN_VEC_FLOAT2.
24124 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
24126         PR rtl-optimization/70023
24127         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
24128         src_regno into account.
24130 2018-02-16  Carl Love  <cel@us.ibm.com>
24132         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
24133         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
24134         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
24135         * config/rs6000/rs6000.c: Remove case statements for
24136         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
24137         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
24138         and P9V_BUILTIN_VEC_VINSERT4B.
24139         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
24140         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
24141         * config/rs6000/vsx.md:
24142         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
24143         vec_insert4b.
24145 2018-02-16  Carl Love  <cel@us.ibm.com>
24147         * config/rs6000/altivec.h: Add builtin names vec_extract4b
24148         vec_insert4b.
24149         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
24150         definitions.
24151         * config/rs6000/rs6000-c.c: Add the definitions for
24152         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
24153         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
24154         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
24155         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
24156         definition for insert4b and define insn *insert3b_internal.
24157         * doc/extend.texi: Add documentation for vec_extract4b.
24159 2018-02-16  Nathan Sidwell  <nathan@acm.org>
24161         * doc/extend.texi (Backwards Compatibility): Mention friend
24162         injection.  Note for-scope is deprecated.
24163         * doc/invoke.texi (-ffriend-injection): Deprecate.
24165 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
24167         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
24168         that moved to I2, also allow destinations that are a paradoxical
24169         subreg (instead of a normal reg).
24171 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
24173         PR target/83831
24174         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
24175         to QImode.
24177 2018-02-16  Richard Biener  <rguenther@suse.de>
24179         PR tree-optimization/84037
24180         PR tree-optimization/84016
24181         PR target/82862
24182         * config/i386/i386.c (ix86_builtin_vectorization_cost):
24183         Adjust vec_construct for the fact we need additional higher latency
24184         128bit inserts for AVX256 and AVX512 vector builds.
24185         (ix86_add_stmt_cost): Scale vector construction cost for
24186         elementwise loads.
24188 2018-02-16  Richard Biener  <rguenther@suse.de>
24190         PR tree-optimization/84417
24191         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
24192         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
24193         (non_rewritable_lvalue_p): Likewise, use poly-ints.
24195 2018-02-16  Martin Liska  <mliska@suse.cz>
24197         PR sanitizer/84307
24198         * internal-fn.def (ASAN_CHECK): Set proper flags.
24199         (ASAN_MARK): Likewise.
24201 2018-02-16  Julia Koval  <julia.koval@intel.com>
24203         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
24204         from PTA_CANNONLAKE.
24206 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
24208         PR target/84272
24209         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
24210         Use ++iter rather than iter++ for std::list iterators.
24211         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
24212         defer deleting them until all nodes in the forest are processed.  Do
24213         free even leaf nodes.  Change to_process into auto_vec.
24215         PR bootstrap/84405
24216         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
24217         * vec.h (vec_default_construct): Use memset instead of placement new
24218         if BROKEN_VALUE_INITIALIZATION is defined.
24219         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
24220         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
24221         is defined.
24223         PR rtl-optimization/83723
24224         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
24225         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
24226         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
24227         recursive calls.
24228         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
24229         callers.
24230         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
24232 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
24234         PR rtl-optimization/81443
24235         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
24236         from inner REGs to paradoxical SUBREGs.
24238 2018-02-16  Richard Biener  <rguenther@suse.de>
24240         PR tree-optimization/84399
24241         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
24242         For operands we can analyze at their definition make sure we can
24243         analyze them at each use as well.
24245 2018-02-16  Richard Biener  <rguenther@suse.de>
24247         PR tree-optimization/84190
24248         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
24249         volatile accesses if the decl isn't volatile.
24251 2018-02-15  Jason Merrill  <jason@redhat.com>
24253         PR c++/84314 - ICE with templates and fastcall attribute.
24254         * attribs.c (build_type_attribute_qual_variant): Don't clobber
24255         TYPE_CANONICAL on an existing type.
24257 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
24259         PR tree-optimization/84383
24260         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
24261         dstoff nor call operand_equal_p if dstbase is NULL.
24263         PR tree-optimization/84334
24264         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
24265         also a CONSTANT_CLASS_P, punt.
24267 2018-02-14  Jim Wilson  <jimw@sifive.com>
24269         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
24270         first SMALL_OPERAND check.  New local min_second_step.  Move assert
24271         to where locals are set.  Add TARGET_RVC support.
24272         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
24274 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
24276         * doc/invoke.texi: Correct -Wformat-overflow code sample.
24278 2018-02-14  Martin Sebor  <msebor@redhat.com>
24280         PR tree-optimization/83698
24281         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
24282         arrays constrain the offset range to their bounds.
24283         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
24284         (builtin_access::overlap): Avoid setting the size of overlap if it's
24285         already been set.
24286         (maybe_diag_overlap): Also consider arrays when deciding what values
24287         of offsets to include in diagnostics.
24289 2018-02-14  Martin Sebor  <msebor@redhat.com>
24291         PR c/84108
24292         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
24293         that correspond to the kind of a declaration.
24295 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
24297         PR target/83984
24298         * config/pa/pa.md: Load address of PIC label using the linkage table
24299         if the label is nonlocal.
24301 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24303         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
24304         warning message if user requests -maltivec=be.
24305         * doc/invoke.texi: Document deprecation of -maltivec=be.
24307 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
24309         PR target/84220
24310         * config/rs6000/rs6000-c.c: Update definitions for
24311         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
24312         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
24314 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
24316         PR target/84239
24317         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
24318         add _get_ssp intrinsics. Remove argument from
24319         __builtin_ia32_rdssp[d|q].
24320         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
24321         * config/i386/i386-builtin.def: Remove argument from
24322         __builtin_ia32_rdssp[d|q].
24323         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
24324         ix86_expand_special_args_builtin for _rdssp[d|q].
24325         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
24326         Clear register before usage.
24327         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
24328         Add documentation for new _get_ssp and _inc_ssp intrinsics.
24330 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
24332         PR tree-optimization/84357
24333         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
24334         operand 1 of an ARRAY_REF too.
24336 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
24338         PR target/83831
24339         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
24340         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
24341         declarations.
24342         (set_of_reg): New struct.
24343         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
24344         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
24345         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
24346         functions.
24347         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
24348         Split into bitclr, bitset, bitinvert patterns if appropriate.
24349         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
24350         use rx_fuse_in_memory_bitop.
24351         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
24352         to named insn, correct maximum insn length.
24354 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
24356         PR target/79242
24357         * machmode.def: Define a complex mode for PARTIAL_INT.
24358         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
24359         MODE_PARTIAL_INT.
24360         * doc/rtl.texi: Document CSPImode.
24361         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
24362         handling.
24363         (msp430_hard_regno_nregs_with_padding): Likewise.
24365 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
24367         PR target/84279
24368         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
24370 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
24372         PR rtl-optimization/84169
24373         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
24374         we generated a parallel as new i3 and we split that to new i2 and i3
24375         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
24376         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
24377         those to i2, not i1.  Partially rewrite this scan code.
24379 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
24381         PR c/82210
24382         * stor-layout.c (place_field): For variable length fields, adjust
24383         offset_align afterwards not just based on the field's alignment,
24384         but also on the size.
24386         PR middle-end/84309
24387         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
24388         of exps and logs in the use_exp2 case.
24390 2018-02-13  Jeff Law  <law@redhat.com>
24392         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
24393         entry for "vector".
24395         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
24396         ARGS as unused.
24398 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
24400         PR debug/84342
24401         PR debug/84319
24402         * common.opt (gas-loc-support, gas-locview-support): New.
24403         (ginline-points, ginternal-reset-location-views): New.
24404         * doc/invoke.texi: Document them.  Use @itemx where intended.
24405         (gvariable-location-views): Adjust.
24406         * target.def (reset_location_view): New.
24407         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
24408         (TARGET_RESET_LOCATION_VIEW): New.
24409         * doc/tm.texi: Rebuilt.
24410         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
24411         (dwarf2out_default_as_locview_support): New.
24412         (output_asm_line_debug_info): Use option variables.
24413         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
24414         (output_loc_list): Likewise.
24415         (add_high_low_attributes): Check option variables.
24416         Don't output entry view attribute in strict mode.
24417         (gen_inlined_subroutine_die): Check option variables.
24418         (dwarf2out_inline_entry): Likewise.
24419         (init_sections_and_labels): Likewise.
24420         (dwarf2out_early_finish): Likewise.
24421         (maybe_reset_location_view): New, from...
24422         (dwarf2out_var_location): ... here.  Call it.
24423         * debug.h (dwarf2out_default_as_loc_support): Declare.
24424         (dwarf2out_default_as_locview_support): Declare.
24425         * hooks.c (hook_int_rtx_insn_0): New.
24426         * hooks.h (hook_int_rtx_insn_0): Declare.
24427         * toplev.c (process_options): Take -gas-loc-support and
24428         -gas-locview-support from dwarf2out.  Enable
24429         -gvariable-location-views by default only with locview
24430         assembler support.  Enable -ginternal-reset-location-views by
24431         default only if the target defines the corresponding hook.
24432         Enable -ginline-points by default if location views are
24433         enabled; force it disabled if statement frontiers are
24434         disabled.
24435         * tree-inline.c (expand_call_inline): Check option variables.
24436         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
24438 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
24440         PR tree-optimization/84321
24441         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
24442         handling.  Also check whether the anti-range contains any values
24443         that satisfy the mask; switch to a VR_RANGE if not.
24445 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
24447         PR sanitizer/84340
24448         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
24450 2018-02-13  Martin Jambor  <mjambor@suse.cz>
24452         PR c++/83990
24453         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
24454         of call statements, also set location of a load to a temporary.
24456 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
24458         * config/rl78/rl78.c (add_vector_labels): New function.
24459         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
24460         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
24461         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
24462         which checks that no arguments are passed.
24463         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
24464         * doc/extend.texi: Documentation for the new attribute.
24466 2018-02-13  Andreas Schwab  <schwab@suse.de>
24468         * config/riscv/linux.h (CPP_SPEC): Define.
24470 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
24472         PR target/84335
24473         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
24474         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
24475         OPTION_MASK_ISA_AES as first argument to def_builtin_const
24476         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
24477         instead of OPTION_MASK_ISA_PCLMUL as first argument to
24478         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
24479         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
24480         temporarily for AES and PCLMUL builtins.
24482         PR tree-optimization/84339
24483         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
24484         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
24485         Formatting fixes.
24487         PR middle-end/84309
24488         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
24489         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
24490         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
24491         inline function.
24492         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
24493         inline function.
24494         * omp-simd-clone.h: New file.
24495         * omp-simd-clone.c: Include omp-simd-clone.h.
24496         (expand_simd_clones): No longer static.
24497         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
24498         cgraph.h and omp-simd-clone.h.
24499         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
24500         (vect_recog_widen_shift_pattern): Formatting fix.
24501         (vect_pattern_recog_1): Don't check optab for calls.
24503         PR target/84336
24504         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
24505         operands[2] into a REG before using gen_lowpart on it.
24507 2018-02-12  Jeff Law  <law@redhat.com>
24509         PR target/83760
24510         * config/sh/sh.c (find_barrier): Consider a sibling call
24511         a barrier as well.
24513         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
24514         successfully back substituting a reg.
24516 2018-02-12  Richard Biener  <rguenther@suse.de>
24518         PR tree-optimization/84037
24519         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
24520         parameter, move visited init to caller.
24521         (vect_slp_analyze_operations): Separate cost from validity
24522         check, initialize visited once for all instances.
24523         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
24524         for all instances.
24525         * tree-vect-stmts.c (vect_model_simple_cost): Make early
24526         out an assert.
24527         (vect_model_promotion_demotion_cost): Likewise.
24528         (vectorizable_bswap): Guard cost modeling with !slp_node
24529         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
24530         SLP stmts.
24531         (vectorizable_call): Likewise.
24532         (vectorizable_conversion): Likewise.
24533         (vectorizable_assignment): Likewise.
24534         (vectorizable_shift): Likewise.
24535         (vectorizable_operation): Likewise.
24536         (vectorizable_store): Likewise.
24537         (vectorizable_load): Likewise.
24538         (vectorizable_condition): Likewise.
24539         (vectorizable_comparison): Likewise.
24541 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
24543         PR sanitizer/84307
24544         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
24545         (ASAN_MARK): Fix fnspec to account for return value, change pointer
24546         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
24548 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
24550         PR middle-end/83665
24551         * params.def (inline-min-speedup): Increase from 8 to 15.
24552         (max-inline-insns-auto): Decrease from 40 to 30.
24553         * ipa-split.c (consider_split): Add some buffer for function to
24554         be considered inlining candidate.
24555         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
24556         default values.
24558 2018-02-12  Richard Biener  <rguenther@suse.de>
24560         PR tree-optimization/84037
24561         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
24562         matched stmts if we cannot swap the non-matched ones.
24564 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
24566         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
24567         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
24568         _mm_maskz_scalef_round_ss): New intrinsics.
24569         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
24570         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
24571         __builtin_ia32_scalefss_round): Remove.
24572         (__builtin_ia32_scalefsd_mask_round,
24573         __builtin_ia32_scalefss_mask_round): New intrinsics.
24574         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
24575         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
24576         ((match_operand:VF_128 2 "<round_nimm_predicate>"
24577         "<round_constraint>")): Changed to ...
24578         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
24579         "<round_scalar_constraint>")): ... this.
24580         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
24581         %0, %1, %2<round_op3>}"): Changed to ...
24582         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
24583         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
24584         %2<round_scalar_mask_op3>}"): ... this.
24585         * config/i386/subst.md (round_scalar_nimm_predicate): New.
24587 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
24589         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
24590         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
24591         (_mm_maskz_sqrt_round_ss): New intrinsics.
24592         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
24593         (__builtin_ia32_sqrtsd_mask_round)
24594         (__builtin_ia32_sqrtss_mask_round): New builtins.
24595         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
24596         (__builtin_ia32_sqrtss_round): Remove.
24597         (__builtin_ia32_sqrtsd_mask_round)
24598         (__builtin_ia32_sqrtss_mask_round): New builtins.
24599         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
24600         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
24601         ((match_operand:VF_128 1 "vector_operand"
24602         "xBm,<round_constraint>")): Changed to ...
24603         ((match_operand:VF_128 1 "vector_operand"
24604         "xBm,<round_scalar_constraint>")): ... this.
24605         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
24606         %0, %2, %<iptr>1<round_op3>}): Changed to ...
24607         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
24608         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
24609         %<iptr>1<round_scalar_mask_op3>}): ... this.
24610         ((set_attr "prefix" "<round_prefix>")): Changed to ...
24611         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
24613 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
24615         PR target/84266
24616         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
24617         Cast vec_cmpeq result to correct type.
24618         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
24619         Cast vec_cmpgt result to correct type.
24621 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
24623         * final.c (final_scan_insn_1): Renamed from...
24624         (final_scan_insn): ... this.  New wrapper, to recover
24625         seen from the outermost call in recursive ones.
24626         * config/sparc/sparc.c (output_return): Drop seen from call.
24627         (output_sibcall): Likewise.
24628         * config/visium/visium.c (output_branch): Likewise.
24630 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
24632         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
24633         function label.
24635 2018-02-10  Alan Modra  <amodra@gmail.com>
24637         PR target/84300
24638         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
24639         Specify LR as an input.
24641 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
24643         PR sanitizer/83987
24644         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
24645         remove_member_access_dummy_vars): New functions.
24646         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
24647         lower_omp_1, execute_lower_omp): Use them.
24649         PR rtl-optimization/84308
24650         * shrink-wrap.c (spread_components): Release todo vector.
24652 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
24654         PR rtl-optimization/57193
24655         * ira-color.c (struct allocno_color_data): Add member
24656         conflict_allocno_hard_prefs.
24657         (update_conflict_allocno_hard_prefs): New.
24658         (bucket_allocno_compare_func): Add a preference based on
24659         conflict_allocno_hard_prefs.
24660         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
24661         (color_allocnos): Remove a dead code.  Initiate
24662         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
24664 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
24666         PR target/84226
24667         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
24668         constraint from =wa to wa.  Avoid a subreg on the output operand,
24669         instead use a pseudo and subreg it in a move.
24670         (p9_xxbrd_<mode>): Changed to ...
24671         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
24672         (p9_xxbrd_v2df): New expander.
24673         (p9_xxbrw_<mode>): Changed to ...
24674         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
24675         (p9_xxbrw_v4sf): New expander.
24677 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
24679         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
24681 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
24683         PR target/83926
24684         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
24685         multiply in 32-bit mode.
24686         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
24687         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
24688         mode.
24690 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
24692         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
24693         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
24694         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
24695         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
24697 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
24699         PR lto/84213
24700         * dwarf2out.c (is_trivial_indirect_ref): New function.
24701         (dwarf2out_late_global_decl): Do not generate a location
24702         attribute for variables that have a non-trivial DECL_VALUE_EXPR
24703         and that are not defined in the current unit.
24705 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
24707         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
24708         instead of a libcall for UNORDERED.
24710 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
24712         PR target/82641
24713         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
24714         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
24716 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24718         PR target/PR84295
24719         * config/s390/s390.c (s390_set_current_function): Invoke
24720         s390_indirect_branch_settings also if fndecl didn't change.
24722 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
24724         * config/rs6000/rs6000.md (blockage): Set length to zero.
24726 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
24728         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
24730 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
24732         PR sanitizer/84285
24733         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
24734         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
24735         -static-lib*san.
24737         PR debug/84252
24738         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
24739         PARALLEL incoming that failed vt_get_decl_and_offset check.
24741         PR middle-end/84237
24742         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
24743         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
24744         TREE_READONLY bit.
24745         (get_variable_section): For decls in named .bss* sections pass true as
24746         second argument to bss_initializer_p.
24748 2018-02-09  Marek Polacek  <polacek@redhat.com>
24749             Jakub Jelinek  <jakub@redhat.com>
24751         PR c++/83659
24752         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
24753         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
24754         Sync some changes from cxx_fold_indirect_ref.
24756 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
24758         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
24759         markers.
24760         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
24761         (BLOCK_INLINE_ENTRY_LABEL): New.
24762         (dwarf2out_var_location): Disregard inline entry markers.
24763         (inline_entry_data): New struct.
24764         (inline_entry_data_hasher): New hashtable type.
24765         (inline_entry_data_hasher::hash): New.
24766         (inline_entry_data_hasher::equal): New.
24767         (inline_entry_data_table): New variable.
24768         (add_high_low_attributes): Add DW_AT_entry_pc and
24769         DW_AT_GNU_entry_view attributes if a pending entry is found
24770         in inline_entry_data_table.  Add old entry_pc attribute only
24771         if debug nonbinding markers are disabled.
24772         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
24773         markers are enabled.
24774         (block_within_block_p, dwarf2out_inline_entry): New.
24775         (dwarf2out_finish): Check that no entries remained in
24776         inline_entry_data_table.
24777         * final.c (reemit_insn_block_notes): Handle inline entry notes.
24778         (final_scan_insn, notice_source_line): Likewise.
24779         (rest_of_clean_state): Skip inline entry markers.
24780         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
24781         markers.
24782         * gimple.c (gimple_build_debug_inline_entry): New.
24783         * gimple.h (enum gimple_debug_subcode): Add
24784         GIMPLE_DEBUG_INLINE_ENTRY.
24785         (gimple_build_debug_inline_entry): Declare.
24786         (gimple_debug_inline_entry_p): New.
24787         (gimple_debug_nonbind_marker_p): Adjust.
24788         * insn-notes.def (INLINE_ENTRY): New.
24789         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
24790         inline entry marker notes.
24791         (print_insn): Likewise.
24792         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
24793         (INSN_DEBUG_MARKER_KIND): Likewise.
24794         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
24795         * tree-inline.c (expand_call_inline): Build and insert
24796         debug_inline_entry stmt.
24797         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
24798         inline entry blocks early, if nonbind markers are enabled.
24799         (dump_scope_block): Dump fragment info.
24800         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
24801         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
24802         (gimple_build_debug_inline_entry): New.
24803         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
24804         Enable/disable inline entry points too.
24805         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
24806         (DEBUG_INSN): Describe inline entry markers.
24808         * common.opt (gvariable-location-views): New.
24809         (gvariable-location-views=incompat5): New.
24810         * config.in: Rebuilt.
24811         * configure: Rebuilt.
24812         * configure.ac: Test assembler for view support.
24813         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
24814         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
24815         * dwarf2out.c (var_loc_view): New typedef.
24816         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
24817         (dwarf2out_locviews_in_attribute): New.
24818         (dwarf2out_locviews_in_loclist): New.
24819         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
24820         (enum dw_line_info_opcode): Add LI_adv_address.
24821         (struct dw_line_info_table): Add view.
24822         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
24823         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
24824         (zero_view_p): New variable.
24825         (ZERO_VIEW_P): New macro.
24826         (output_asm_line_debug_info): New.
24827         (struct var_loc_node): Add view.
24828         (add_AT_view_list, AT_loc_list): New.
24829         (add_var_loc_to_decl): Add view param.  Test it against last.
24830         (new_loc_list): Add view params.  Record them.
24831         (AT_loc_list_ptr): Handle loc and view lists.
24832         (view_list_to_loc_list_val_node): New.
24833         (print_dw_val): Handle dw_val_class_view_list.
24834         (size_of_die): Likewise.
24835         (value_format): Likewise.
24836         (loc_list_has_views): New.
24837         (gen_llsym): Set vl_symbol too.
24838         (maybe_gen_llsym, skip_loc_list_entry): New.
24839         (dwarf2out_maybe_output_loclist_view_pair): New.
24840         (output_loc_list): Output view list or entries too.
24841         (output_view_list_offset): New.
24842         (output_die): Handle dw_val_class_view_list.
24843         (output_dwarf_version): New.
24844         (output_compilation_unit_header): Use it.
24845         (output_skeleton_debug_sections): Likewise.
24846         (output_rnglists, output_line_info): Likewise.
24847         (output_pubnames, output_aranges): Update version comments.
24848         (output_one_line_info_table): Output view numbers in asm comments.
24849         (dw_loc_list): Determine current endview, pass it to new_loc_list.
24850         Call maybe_gen_llsym.
24851         (loc_list_from_tree_1): Adjust.
24852         (add_AT_location_description): Create view list attribute if
24853         needed, check it's absent otherwise.
24854         (convert_cfa_to_fb_loc_list): Adjust.
24855         (maybe_emit_file): Call output_asm_line_debug_info for test.
24856         (dwarf2out_var_location): Reset views as needed.  Precompute
24857         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
24858         attribute.  Set view.
24859         (new_line_info_table): Reset next view.
24860         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
24861         (dwarf2out_source_line): Likewise.  Output view resets and labels to
24862         the assembler, or select appropriate line info opcodes.
24863         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
24864         (optimize_string_length): Catch it.  Adjust.
24865         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
24866         dw_val_class_view_list, and remove it if no longer needed.
24867         (hash_loc_list): Hash view numbers.
24868         (loc_list_hasher::equal): Compare them.
24869         (optimize_location_lists): Check whether a view list symbol is
24870         needed, and whether the locview attribute is present, and
24871         whether they match.  Remove the locview attribute if no longer
24872         needed.
24873         (index_location_lists): Call skip_loc_list_entry for test.
24874         (dwarf2out_finish): Call output_asm_line_debug_info for test.
24875         Use output_dwarf_version.
24876         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
24877         (struct dw_val_node): Add val_view_list.
24878         * final.c (SEEN_NEXT_VIEW): New.
24879         (set_next_view_needed): New.
24880         (clear_next_view_needed): New.
24881         (maybe_output_next_view): New.
24882         (final_start_function): Rename to...
24883         (final_start_function_1): ... this.  Take pointer to FIRST,
24884         add SEEN parameter.  Emit param bindings in the initial view.
24885         (final_start_function): Reintroduce SEEN-less interface.
24886         (final): Rename to...
24887         (final_1): ... this.  Take SEEN parameter.  Output final pending
24888         next view at the end.
24889         (final): Reintroduce seen-less interface.
24890         (final_scan_insn): Output pending next view before switching
24891         sections or ending a block.  Mark the next view as needed when
24892         outputting variable locations.  Notify debug backend of section
24893         changes, and of location view changes.
24894         (rest_of_handle_final): Adjust.
24895         * toplev.c (process_options): Autodetect value for debug variable
24896         location views option.  Warn on incompat5 without -gdwarf-5.
24897         * doc/invoke.texi (gvariable-location-views): New.
24898         (gvariable-location-views=incompat5): New.
24899         (gno-variable-location-views): New.
24901 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
24903         PR tree-optimization/84136
24904         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
24905         that the result of find_edge is non-NULL.
24907 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
24909         PR target/83008
24910         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
24911         storing integer register in SImode.  Fix cost of 256 and 512
24912         byte aligned SSE register store.
24914 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
24916         * config/i386/i386.c (ix86_multiplication_cost): Fix
24917         multiplication cost for TARGET_AVX512DQ.
24919 2018-02-08  Marek Polacek  <polacek@redhat.com>
24921         PR tree-optimization/84238
24922         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
24923         get_range_strlen.
24925 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
24927         PR tree-optimization/84265
24928         * tree-vect-stmts.c (vectorizable_store): Don't treat
24929         VMAT_CONTIGUOUS accesses as grouped.
24930         (vectorizable_load): Likewise.
24932 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
24934         PR tree-optimization/81635
24935         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
24936         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
24937         (test_round_for_mask): New functions.
24938         (wide_int_cc_tests): Call test_round_for_mask.
24939         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
24940         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
24941         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
24942         range returned by get_range_info.
24944 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
24946         PR ipa/81360
24947         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
24948         * symtab.c: Include builtins.h
24949         (symtab_node::output_to_lto_symbol_table_p): Move here
24950         from lto-streamer-out.c:output_symbol_p.
24951         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
24952         (output_symbol_p): Move all logic to symtab.c
24953         (produce_symtab): Update.
24955 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24957         * config/s390/s390-opts.h (enum indirect_branch): Define.
24958         * config/s390/s390-protos.h (s390_return_addr_from_memory)
24959         (s390_indirect_branch_via_thunk)
24960         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
24961         (enum s390_indirect_branch_type): Define.
24962         * config/s390/s390.c (struct s390_frame_layout, struct
24963         machine_function): Remove.
24964         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
24965         (indirect_branch_table_label_no, indirect_branch_table_name):
24966         Define variables.
24967         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
24968         (enum s390_indirect_branch_option): Define.
24969         (s390_return_addr_from_memory): New function.
24970         (s390_handle_string_attribute): New function.
24971         (s390_attribute_table): Add new attribute handler.
24972         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
24973         (s390_indirect_branch_via_thunk): New function.
24974         (s390_indirect_branch_via_inline_thunk): New function.
24975         (s390_function_ok_for_sibcall): When jumping via thunk disallow
24976         sibling call optimization for non z10 compiles.
24977         (s390_emit_call): Force indirect branch target to be a single
24978         register.  Add r1 clobber for non-z10 compiles.
24979         (s390_emit_epilogue): Emit return jump via return_use expander.
24980         (s390_reorg): Handle JUMP_INSNs as execute targets.
24981         (s390_option_override_internal): Perform validity checks for the
24982         new command line options.
24983         (s390_indirect_branch_attrvalue): New function.
24984         (s390_indirect_branch_settings): New function.
24985         (s390_set_current_function): Invoke s390_indirect_branch_settings.
24986         (s390_output_indirect_thunk_function):  New function.
24987         (s390_code_end): Implement target hook.
24988         (s390_case_values_threshold): Implement target hook.
24989         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
24990         macros.
24991         * config/s390/s390.h (struct s390_frame_layout)
24992         (struct machine_function): Move here from s390.c.
24993         (TARGET_INDIRECT_BRANCH_NOBP_RET)
24994         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
24995         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
24996         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
24997         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
24998         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
24999         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
25000         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
25001         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
25002         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
25003         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
25004         (mnemonic attribute): Add values which aren't recognized
25005         automatically.
25006         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
25007         pattern for branch conversion.  Fix mnemonic attribute.
25008         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
25009         indirect branch via thunk if requested.
25010         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
25011         ("*indirect_jump"): Disable for branch conversion using out of
25012         line thunks.
25013         ("indirect_jump_via_thunk<mode>_z10")
25014         ("indirect_jump_via_thunk<mode>")
25015         ("indirect_jump_via_inlinethunk<mode>_z10")
25016         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
25017         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
25018         ("casesi_jump_via_inlinethunk<mode>_z10")
25019         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
25020         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
25021         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
25022         ("*indirect2_jump"): Disable for branch conversion.
25023         ("casesi_jump"): Turn into expander and expand patterns for branch
25024         conversion.
25025         ("return_use"): New expander.
25026         ("*return"): Emit return via thunk and rename it to ...
25027         ("*return<mode>"): ... this one.
25028         * config/s390/s390.opt: Add new options and and enum for the
25029         option values.
25031 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
25033         * lra-constraints.c (match_reload): Unconditionally use
25034         gen_lowpart_SUBREG, rather than selecting between that
25035         and equivalent gen_rtx_SUBREG code.
25037 2018-02-08  Richard Biener  <rguenther@suse.de>
25039         PR tree-optimization/84233
25040         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
25041         changed flag instead of boguously re-using phi_inserted.
25043 2018-02-08  Martin Jambor  <mjambor@suse.cz>
25045         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
25046         static local variables.
25048 2018-02-08  Richard Biener  <rguenther@suse.de>
25050         PR tree-optimization/84278
25051         * tree-vect-stmts.c (vectorizable_store): When looking for
25052         smaller vector types to perform grouped strided loads/stores
25053         make sure the mode is supported by the target.
25054         (vectorizable_load): Likewise.
25056 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
25058         * config/aarch64/aarch64.c (aarch64_components_for_bb):
25059         Increase LDP/STP opportunities by adding adjacent callee-saves.
25061 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
25063         PR rtl-optimization/84068
25064         PR rtl-optimization/83459
25065         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
25067 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
25069         PR tree-optimization/84224
25070         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
25071         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
25072         non-zero arguments.
25074 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
25076         PR target/84113
25077         * config/rs6000/altivec.md (*restore_world): Remove LR use.
25078         * config/rs6000/predicates.md (restore_world_operation): Adjust op
25079         count, remove one USE.
25081 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
25083         * doc/install.texi (Configuration): Document the
25084         --with-long-double-format={ibm,ieee} PowerPC configuration
25085         options.
25087         PR target/84154
25088         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
25089         Convert from define_expand to be define_insn_and_split.  Rework
25090         float/double/_Float128 conversions to QI/HI/SImode to work with
25091         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
25092         conversions to QI/HImode types did a store and then a load to
25093         truncate the value.  For conversions to VSX registers, don't split
25094         the insn, instead emit the code directly.  Use the code iterator
25095         any_fix to combine signed and unsigned conversions.
25096         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
25097         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
25098         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
25099         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
25100         (fix_<mode>di2_hw): Likewise.
25101         (fixuns_<mode>di2_hw): Likewise.
25102         (fix_<mode>si2_hw): Likewise.
25103         (fixuns_<mode>si2_hw): Likewise.
25104         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
25105         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
25106         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
25107         fix<uns>_trunc<SFDF:mode>si2_p8.
25108         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
25109         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
25110         (fix<uns>_<mode>_mem): Likewise.
25111         (fctiw<u>z_<mode>_mem): Likewise.
25112         (fix<uns>_<mode>_mem): Likewise.
25113         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
25114         the register allocator from doing a direct move to the GPRs to do
25115         a store, and instead use the ISA 3.0 store byte/half-word from
25116         vector register instruction.  For IEEE 128-bit floating point,
25117         also optimize stores of 32-bit ints.
25118         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
25120 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
25122         * genextract.c (push_pathstr_operand): New function to support
25123         [a-zA-Z].
25124         (walk_rtx): Call push_pathstr_operand.
25125         (print_path): Support [a-zA-Z].
25127 2018-02-07  Richard Biener  <rguenther@suse.de>
25129         PR tree-optimization/84037
25130         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
25131         (cse_and_gimplify_to_preheader): Declare.
25132         (vect_get_place_in_interleaving_chain): Likewise.
25133         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25134         ivexpr_map.
25135         (_loop_vec_info::~_loop_vec_info): Delete it.
25136         (cse_and_gimplify_to_preheader): New function.
25137         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
25138         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
25139         (vectorizable_load): Likewise.  For grouped stores always base
25140         the IV on the first element.
25141         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
25142         condition before gimplifying.
25144 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
25146         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
25147         *DIV_EXPR and *MOD_EXPR.
25149 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
25151         PR target/84248
25152         * config/i386/i386.c (ix86_option_override_internal): Mask out
25153         the CF_SET bit when checking -fcf-protection.
25155 2018-02-07  Tom de Vries  <tom@codesourcery.com>
25157         PR libgomp/84217
25158         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
25159         enough.
25161 2018-02-07  Richard Biener  <rguenther@suse.de>
25163         PR tree-optimization/84204
25164         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
25165         this place.
25167         PR tree-optimization/84205
25168         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
25169         special-case isl_ast_op_zdiv_r.
25171         PR tree-optimization/84223
25172         * graphite-scop-detection.c (gather_bbs::before_dom_children):
25173         Only add conditions from within the region.
25174         (gather_bbs::after_dom_children): Adjust.
25176 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
25178         PR target/84209
25179         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
25180         * config/avr/avr.md: Only post-reload split REG-REG moves if
25181         either register is GENERAL_REG_P.
25183 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
25185         PR tree-optimization/84235
25186         * tree-ssa-scopedtables.c
25187         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
25188         if the subtraction is performed in floating point type where NaNs are
25189         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
25190         build 1.  Formatting fix.
25192 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
25194         PR target/84146
25195         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
25196         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
25197         and skip it regardless of bb boundaries.  Use CALL_P macro,
25198         don't test INSN_P (insn) together with CALL_P or JUMP_P check
25199         unnecessarily, formatting fix.
25201 2018-02-06  Michael Collison  <michael.collison@arm.com>
25203         * config/arm/thumb2.md:
25204         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
25205         (*thumb_mov_notscc): Ditto.
25207 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
25209         PR target/84154
25210         * config/rs6000/rs6000.md (su code attribute): Use "u" for
25211         unsigned_fix, not "s".
25213 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25215         * configure.ac (gcc_fn_eh_frame_ro): New function.
25216         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
25217         correct .eh_frame permissions.
25218         * configure: Regenerate.
25220 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
25222         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
25223         irrelevant options.
25225 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25227         * config/rs6000/rs6000.c (rs6000_option_override_internal):
25228         Display warning message for -mno-speculate-indirect-jumps.
25230 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
25232         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
25233         Undocumented.
25234         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
25236 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
25238         PR tree-optimization/84225
25239         * tree-eh.c (find_trapping_overflow): Only call
25240         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
25242 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
25244         PR target/84145
25245         * config/i386/i386.c: Reimplement the check of possible options
25246         -mibt/-mshstk conbination. Change error messages.
25247         * doc/invoke.texi: Fix a typo: remove extra '='.
25249 2018-02-06  Marek Polacek  <polacek@redhat.com>
25251         PR tree-optimization/84228
25252         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
25254 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
25256         PR target/82641
25257         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
25258         emitted arch directives.
25259         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
25260         __ARM_FEATURE_COPROC before changing architectures.
25262 2018-02-06  Richard Biener  <rguenther@suse.de>
25264         * config/i386/i386.c (print_reg): Fix typo.
25265         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
25267 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
25269         * configure: Regenerate.
25271 2018-02-05  Martin Sebor  <msebor@redhat.com>
25273         PR tree-optimization/83369
25274         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
25275         inlining context.
25277 2018-02-05  Martin Liska  <mliska@suse.cz>
25279         * doc/invoke.texi: Cherry-pick upstream r323995.
25281 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
25283         * ira.c (ira_init_register_move_cost): Adjust comment.
25285 2018-02-05  Martin Liska  <mliska@suse.cz>
25287         PR gcov-profile/84137
25288         * doc/gcov.texi: Fix typo in documentation.
25290 2018-02-05  Martin Liska  <mliska@suse.cz>
25292         PR gcov-profile/83879
25293         * doc/gcov.texi: Document necessity of --dynamic-list-data when
25294         using dlopen functionality.
25296 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
25298         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
25299         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
25300         _mm_maskz_range_ss, _mm_mask_range_round_ss,
25301         _mm_maskz_range_round_ss): New intrinsics.
25302         (__builtin_ia32_rangesd128_round)
25303         (__builtin_ia32_rangess128_round): Remove.
25304         (__builtin_ia32_rangesd128_mask_round,
25305         __builtin_ia32_rangess128_mask_round): New builtins.
25306         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
25307         __builtin_ia32_rangess128_round): Remove.
25308         (__builtin_ia32_rangesd128_mask_round,
25309         __builtin_ia32_rangess128_mask_round): New builtins.
25310         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
25311         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
25312         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
25313         "<round_saeonly_constraint>")): Changed to ...
25314         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
25315         "<round_saeonly_scalar_constraint>")): ... this.
25316         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25317         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
25318         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
25319         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
25320         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
25322 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
25324         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
25325         options.
25326         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
25327         Remove all values except native, 8540 and 8548.
25329 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
25331         * config/i386/i386.c (ix86_output_function_return): Pass
25332         INVALID_REGNUM, instead of -1, as invalid register number to
25333         indirect_thunk_name and output_indirect_thunk.
25335 2018-02-02  Julia Koval  <julia.koval@intel.com>
25337         * config.gcc: Add -march=icelake.
25338         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
25339         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
25340         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
25341         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
25342         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
25343         (processor_target_table): Add icelake.
25344         (ix86_option_override_internal): Handle new PTAs.
25345         (get_builtin_code_for_version): Handle icelake.
25346         (M_INTEL_COREI7_ICELAKE): New.
25347         (fold_builtin_cpu): Handle icelake.
25348         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
25349         * doc/invoke.texi: Add -march=icelake.
25351 2018-02-02  Julia Koval  <julia.koval@intel.com>
25353         * config/i386/i386.c (ix86_option_override_internal): Change flags type
25354         to wide_int_bitmask.
25355         * wide-int-bitmask.h: New.
25357 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
25359         PR target/84066
25360         * config/i386/i386.md: Replace Pmode with word_mode in
25361         builtin_setjmp_setup and builtin_longjmp to support x32.
25363 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
25365         PR target/56010
25366         PR target/83743
25367         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
25368         #include "opts.h".
25369         (rs6000_supported_cpu_names): New static variable.
25370         (linux_cpu_translation_table): Likewise.
25371         (elf_platform) <cpu>: Define new static variable and use it.
25372         Translate kernel AT_PLATFORM name to canonical name if needed.
25373         Error if platform name is unknown.
25375 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
25377         PR target/84089
25378         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
25380 2018-02-01  Jeff Law  <law@redhat.com>
25382         PR target/84128
25383         * config/i386/i386.c (release_scratch_register_on_entry): Add new
25384         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
25385         the scratch if RELEASE_VIA_POP is false.
25386         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
25387         If we have to save a temporary register, decrement SIZE appropriately.
25388         Pass new arguments to release_scratch_register_on_entry.
25389         (ix86_adjust_stack_and_probe): Likewise.
25390         (ix86_emit_probe_stack_range): Pass new arguments to
25391         release_scratch_register_on_entry.
25393 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
25395         PR rtl-optimization/84157
25396         * combine.c (change_zero_ext): Use REG_P predicate in
25397         front of HARD_REGISTER_P predicate.
25399 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
25401         * config/avr/avr.c (avr_option_override): Move disabling of
25402         -fdelete-null-pointer-checks to...
25403         * common/config/avr/avr-common.c (avr_option_optimization_table):
25404         ...here.
25406 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25408         PR tree-optimization/81635
25409         * tree-data-ref.c (split_constant_offset_1): For types that
25410         wrap on overflow, try to use range info to prove that wrapping
25411         cannot occur.
25413 2018-02-01  Renlin Li  <renlin.li@arm.com>
25415         PR target/83370
25416         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
25417         TAILCALL_ADDR_REGS.
25418         (aarch64_register_move_cost): Likewise.
25419         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
25420         TAILCALL_ADDR_REGS.
25421         (REG_CLASS_NAMES): Likewise.
25422         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
25423         TAILCALL_ADDR_REGS. Remove IP registers.
25424         * config/aarch64/aarch64.md (Ucs): Update register constraint.
25426 2018-02-01  Richard Biener  <rguenther@suse.de>
25428         * domwalk.h (dom_walker::dom_walker): Add additional constructor
25429         for specifying RPO order and allow NULL for that.
25430         * domwalk.c (dom_walker::dom_walker): Likewise.
25431         (dom_walker::walk): Handle NULL RPO order.
25432         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
25433         in RPO order.
25434         (rewrite_update_dom_walker): Likewise.
25435         (mark_def_dom_walker): Likewise.
25437 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25439         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
25440         (aarch64_maybe_expand_sve_subreg_move): Declare.
25441         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
25442         * config/aarch64/predicates.md (aarch64_any_register_operand): New
25443         predicate.
25444         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
25445         that are semantically a reverse operation.
25446         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
25447         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
25448         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
25449         functions.
25450         (aarch64_can_change_mode_class): For big-endian, forbid changes
25451         between two SVE modes if they have different element sizes.
25453 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25455         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
25456         the TImode handling for big-endian targets.
25458 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25460         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
25461         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
25462         not just bytes.
25463         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
25464         Remove BSWAP handing for big-endian targets and use the form of
25465         LD1RQ appropariate for the mode.
25467 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25469         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
25470         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
25471         duplicated element.
25473 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
25475         PR tearget/83845
25476         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
25477         check for operands that need to go through aarch64_sve_reload_be.
25479 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
25481         PR tree-optimization/81661
25482         PR tree-optimization/84117
25483         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
25484         * tree-eh.c: Include gimplify.h.
25485         (find_trapping_overflow, replace_trapping_overflow,
25486         rewrite_to_non_trapping_overflow): New functions.
25487         * tree-vect-loop.c: Include tree-eh.h.
25488         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
25489         * tree-data-ref.c: Include tree-eh.h.
25490         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
25492 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
25494         PR rtl-optimization/84123
25495         * combine.c (change_zero_ext): Check if hard register satisfies
25496         can_change_dest_mode before calling gen_lowpart_SUBREG.
25498 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
25500         PR target/82444
25501         * ira.c (ira_init_register_move_cost): Remove assert.
25503 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
25505         PR rtl-optimization/84071
25506         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
25507         * doc/tm.texi: Regenerate.
25509 2018-01-31  Richard Biener  <rguenther@suse.de>
25511         PR tree-optimization/84132
25512         * tree-data-ref.c (analyze_miv_subscript): Properly
25513         check whether evolution_function_is_affine_multivariate_p
25514         before calling gcd_of_steps_may_divide_p.
25516 2018-01-31  Julia Koval  <julia.koval@intel.com>
25518         PR target/83618
25519         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
25520         * config/i386/i386.md (rdpid_rex64) New.
25521         (rdpid): Make 32bit only.
25523 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
25525         PR lto/84105
25526         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
25527         an IDENTIFIER_NODE for FUNCTION_TYPE's.
25529 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
25531         Revert
25532         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
25534         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
25536 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
25538         PR rtl-optimization/84071
25539         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
25540         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
25542 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
25544         * config/arc/arc.c (arc_handle_aux_attribute): New function.
25545         (arc_attribute_table): Add 'aux' attribute.
25546         (arc_in_small_data_p): Consider aux like variables.
25547         (arc_is_aux_reg_p): New function.
25548         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
25549         (arc_get_aux_arg): New function.
25550         (prepare_move_operands): Handle aux-register access.
25551         (arc_handle_aux_attribute): New function.
25552         * doc/extend.texi (ARC Variable attributes): Add subsection.
25554 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
25556         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
25557         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
25558         (arc_attribute_table): Add 'uncached' attribute.
25559         (arc_print_operand): Print '.di' flag for uncached memory
25560         accesses.
25561         (arc_in_small_data_p): Do not consider for small data the uncached
25562         types.
25563         (arc_is_uncached_mem_p): New function.
25564         * config/arc/predicates.md (compact_store_memory_operand): Check
25565         for uncached memory accesses.
25566         (nonvol_nonimm_operand): Likewise.
25567         * doc/extend.texi (ARC Type Attribute): New subsection.
25569 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
25571         PR c/84100
25572         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
25573         falign-loops=): Add Optimization flag.
25575 2018-01-30  Jeff Law  <law@redhat.com>
25577         PR target/84064
25578         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
25579         INT_REGISTERS_SAVED.  Check it prior to calling
25580         get_scratch_register_on_entry.
25581         (ix86_adjust_stack_and_probe): Similarly.
25582         (ix86_emit_probe_stack_range): Similarly.
25583         (ix86_expand_prologue): Corresponding changes.
25585 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25587         PR target/40411
25588         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
25589         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
25591 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
25593         PR target/84112
25594         * lra-constraints.c (curr_insn_transform): Process AND in the
25595         address.
25597 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
25599         PR rtl-optimization/83986
25600         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
25601         dependence against last_pending_memory_flush in addition to
25602         pending_jump_insns.
25604 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
25606         PR tree-optimization/81611
25607         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
25608         copies.
25610 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25612         PR target/83758
25613         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
25614         a reg rtx.
25616 2018-01-30  Richard Biener  <rguenther@suse.de>
25617             Jakub Jelinek  <jakub@redhat.com>
25619         PR tree-optimization/84111
25620         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
25621         inner loops added during recursion, as they don't have up-to-date
25622         SSA form.
25624 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
25626         PR ipa/81360
25627         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
25628         (can_inline_edge_by_limits_p): ... here.
25629         (can_early_inline_edge_p, check_callers,
25630         update_caller_keys, update_callee_keys, recursive_inlining,
25631         add_new_edges_to_heap, speculation_useful_p,
25632         inline_small_functions,
25633         inline_small_functions, flatten_function,
25634         inline_to_all_callers_1): Update.
25636 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
25638         * profile-count.c (profile_count::combine_with_ipa_count): Handle
25639         zeros correctly.
25641 2018-01-30  Richard Biener  <rguenther@suse.de>
25643         PR tree-optimization/83008
25644         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
25645         invariant and constant vector uses in stmts when they need
25646         more than one stmt.
25648 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25650         PR bootstrap/84017
25651         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
25652         * configure: Regenerate.
25654 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
25656         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
25657         pattern.
25658         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
25659         Use gen_rtx_REG rather than gen_lowpart.
25661 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
25663         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
25664         rather than 0 when creating partial subregs.
25666 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
25668         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
25669         of usage.
25671 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
25673         PR target/81550
25674         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
25675         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
25676         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
25677         flags.  This restores the settings used before the 2017-07-24.
25678         Turning off pre increment/decrement/modify allows IVOPTS to
25679         optimize DF/SF loops where the index is an int.
25681 2018-01-29  Richard Biener  <rguenther@suse.de>
25682             Kelvin Nilsen  <kelvin@gcc.gnu.org>
25684         PR bootstrap/80867
25685         * tree-vect-stmts.c (vectorizable_call): Don't call
25686         targetm.vectorize_builtin_md_vectorized_function if callee is
25687         NULL.
25689 2018-01-22  Carl Love  <cel@us.ibm.com>
25691         * doc/extend.tex: Fix typo in second arg in
25692         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
25694 2018-01-29  Richard Biener  <rguenther@suse.de>
25696         PR tree-optimization/84086
25697         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
25698         (flush_ssaname_freelist): When SSA names were released reset
25699         the SCEV hash table.
25701 2018-01-29  Richard Biener  <rguenther@suse.de>
25703         PR tree-optimization/84057
25704         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
25705         removed paths when removing edges.
25707 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
25709         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
25710         -mfunction-return=@var{choice}.
25712 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
25714         PR diagnostic/84034
25715         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
25716         Handle CR like TAB.
25717         (layout::print_source_line): Likewise.
25718         (test_get_line_width_without_trailing_whitespace): Add test cases.
25720 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
25722         PR middle-end/84040
25723         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
25724         debug insns.
25726 2018-01-26  Jim Wilson  <jimw@sifive.com>
25728         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
25730         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
25731         specified.
25733 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25735         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
25736         and CMP + SUB-immediate -> SUBS.
25738 2018-01-26  Martin Sebor  <msebor@redhat.com>
25740         PR tree-optimization/83896
25741         * tree-ssa-strlen.c (get_string_len): Rename...
25742         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
25743         Avoid assuming length is constant.
25744         (handle_char_store): Use HOST_WIDE_INT for string length.
25746 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
25748         PR target/81763
25749         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
25750         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
25752 2018-01-26  Richard Biener  <rguenther@suse.de>
25754         PR rtl-optimization/84003
25755         * dse.c (record_store): Only record redundant stores when
25756         the earlier store aliases at least all accesses the later one does.
25758 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
25760         PR rtl-optimization/83985
25761         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
25762         REG_CFA_RESTORE insns.
25763         (delete_unmarked_insns): Don't ignore separate shrink wrapping
25764         REG_CFA_RESTORE insns here.
25766         PR c/83989
25767         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
25768         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
25770 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
25772         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
25773         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
25774         (arc_init): Likewise.
25775         (arc_override_options): Likewise.
25776         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
25777         value.
25778         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
25779         support.
25780         * config/arc/arc.h (TARGET_DBNZ): Define.
25781         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
25782         properly set the tune attribute.
25783         (dbnz): Use TARGET_DBNZ guard.
25784         * config/arc/arc.opt (mtune): Add core3 option.
25786 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
25788         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
25789         recognize new pic like addresses.
25790         (arc_delegitimize_address): Clean up.
25792 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
25794         * config/arc/arc-arches.def: Option mrf16 valid for all
25795         architectures.
25796         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
25797         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
25798         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
25799         * config/arc/arc-tables.opt: Regenerate.
25800         * config/arc/arc.c (arc_conditional_register_usage): Handle
25801         reduced register file case.
25802         (arc_file_start): Set must have build attributes.
25803         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
25804         mrf16 option value.
25805         * config/arc/arc.opt (mrf16): Add new option.
25806         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
25807         * config/arc/genmultilib.awk: Handle new mrf16 option.
25808         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
25809         * config/arc/t-multilib: Regenerate.
25810         * doc/invoke.texi (ARC Options): Document mrf16 option.
25812 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
25814         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
25815         * config/arc/arc.c (arc_handle_secure_attribute): New function.
25816         (arc_attribute_table): Add 'secure_call' attribute.
25817         (arc_print_operand): Print secure call operand.
25818         (arc_function_ok_for_sibcall): Don't optimize tail calls when
25819         secure.
25820         (arc_is_secure_call_p): New function.  * config/arc/arc.md
25821         (call_i): Add support for sjli instruction.
25822         (call_value_i): Likewise.
25823         * config/arc/constraints.md (Csc): New constraint.
25825 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
25826             John Eric Martin  <John.Martin@emmicro-us.com>
25828         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
25829         * config/arc/arc.c (_arc_jli_section): New struct.
25830         (arc_jli_section): New type.
25831         (rc_jli_sections): New static variable.
25832         (arc_handle_jli_attribute): New function.
25833         (arc_attribute_table): Add jli_always and jli_fixed attribute.
25834         (arc_file_end): New function.
25835         (TARGET_ASM_FILE_END): Define.
25836         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
25837         (arc_add_jli_section): New function.
25838         (jli_call_scan): Likewise.
25839         (arc_reorg): Call jli_call_scan.
25840         (arc_output_addsi): Remove 'S' from printing asm operand.
25841         (arc_is_jli_call_p): New function.
25842         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
25843         operand.
25844         (movhi_insn): Likewise.
25845         (movsi_insn): Likewise.
25846         (movsi_set_cc_insn): Likewise.
25847         (loadqi_update): Likewise.
25848         (load_zeroextendqisi_update): Likewise.
25849         (load_signextendqisi_update): Likewise.
25850         (loadhi_update): Likewise.
25851         (load_zeroextendhisi_update): Likewise.
25852         (load_signextendhisi_update): Likewise.
25853         (loadsi_update): Likewise.
25854         (loadsf_update): Likewise.
25855         (movsicc_insn): Likewise.
25856         (bset_insn): Likewise.
25857         (bxor_insn): Likewise.
25858         (bclr_insn): Likewise.
25859         (bmsk_insn): Likewise.
25860         (bicsi3_insn): Likewise.
25861         (cmpsi_cc_c_insn): Likewise.
25862         (movsi_ne): Likewise.
25863         (movsi_cond_exec): Likewise.
25864         (clrsbsi2): Likewise.
25865         (norm_f): Likewise.
25866         (normw): Likewise.
25867         (swap): Likewise.
25868         (divaw): Likewise.
25869         (flag): Likewise.
25870         (sr): Likewise.
25871         (kflag): Likewise.
25872         (ffs): Likewise.
25873         (ffs_f): Likewise.
25874         (fls): Likewise.
25875         (call_i): Remove 'S' asm letter, add jli instruction.
25876         (call_value_i): Likewise.
25877         * config/arc/arc.op (mjli-always): New option.
25878         * config/arc/constraints.md (Cji): New constraint.
25879         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
25880         operand.
25881         (subsf3_fpx): Likewise.
25882         (mulsf3_fpx): Likewise.
25883         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
25884         asm operand.
25885         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
25886         function attrbutes.
25887         * doc/invoke.texi (ARC): Document mjli-always option.
25889 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
25891         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
25892         avoid addition with 0 and use incw and decw where possible.
25894 2018-01-26  Richard Biener  <rguenther@suse.de>
25896         PR tree-optimization/81082
25897         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
25898         association if it requires casting to unsigned.
25899         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
25900         from fold_plusminus_mult_expr to catch important cases late when
25901         range info is available.
25903 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25905         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
25906         * configure.ac (hidden_linkonce): New test.
25907         * configure: Regenerate.
25908         * config.in: Regenerate.
25910 2018-01-26  Julia Koval  <julia.koval@intel.com>
25912         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
25913         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
25914         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
25915         _mm_mask_bitshuffle_epi64_mask): Fix type.
25916         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
25917         USI_FTYPE_V4DI_V4DI_USI): Remove.
25918         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
25919         __builtin_ia32_vpshufbitqmb256_mask,
25920         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
25921         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
25922         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
25924 2018-01-26  Alan Modra  <amodra@gmail.com>
25926         PR target/84033
25927         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
25928         UNSPEC_VBPERMQ.  Sort other unspecs.
25930 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
25932         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
25934 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
25936         PR middle-end/83055
25937         * predict.c (drop_profile): Do not push/pop cfun; update also
25938         node->count.
25939         (handle_missing_profiles): Fix logic looking for zero profiles.
25941 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
25943         PR middle-end/83977
25944         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
25945         on functions with #pragma omp declare simd or functions with simd
25946         attribute.
25947         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
25948         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
25949         Remove trailing \n from warning_at calls.
25951 2018-01-25  Tom de Vries  <tom@codesourcery.com>
25953         PR target/84028
25954         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
25955         for neutered workers.
25957 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
25959         PR target/68467
25960         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
25961         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
25963 2018-01-24  Jeff Law  <law@redhat.com>
25965         PR target/83994
25966         * i386.c (get_probe_interval): Move to earlier point.
25967         (ix86_compute_frame_layout): If -fstack-clash-protection and
25968         the frame is larger than the probe interval, then use pushes
25969         to save registers rather than reg->mem moves.
25970         (ix86_expand_prologue): Remove conditional for int_registers_saved
25971         assertion.
25973 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
25975         PR target/84014
25976         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
25977         min/max for never referenced object.
25979 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
25981         PR middle-end/83977
25982         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
25983         here.
25984         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
25985         attributes from DECL_ATTRIBUTES (decl) without affecting
25986         DECL_ATTRIBUTES (current_function_decl).
25987         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
25988         functions with non-NULL DECL_ABSTRACT_ORIGIN.
25990 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
25992         PR tree-optimization/83979
25993         * fold-const.c (fold_comparison): Use constant_boolean_node
25994         instead of boolean_{true,false}_node.
25996 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
25998         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
25999         with zero counts.
26001 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26003         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
26004         Simplify the clause that sets the length attribute.
26005         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
26006         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
26007         clause that sets the length attribute.
26008         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
26010 2018-01-24  Tom de Vries  <tom@codesourcery.com>
26012         PR target/83589
26013         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
26014         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
26015         Add strict parameter.
26016         (prevent_branch_around_nothing): Insert dummy insn between branch to
26017         label and label with no ptx insn inbetween.
26018         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
26020 2018-01-24  Tom de Vries  <tom@codesourcery.com>
26022         PR target/81352
26023         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
26024         for neutered threads in warp.
26025         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
26027 2018-01-24  Richard Biener  <rguenther@suse.de>
26029         PR tree-optimization/83176
26030         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
26031         operands.
26033 2018-01-24  Richard Biener  <rguenther@suse.de>
26035         PR tree-optimization/82819
26036         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
26037         code generating pluses that are no-ops in the target precision.
26039 2018-01-24  Richard Biener  <rguenther@suse.de>
26041         PR middle-end/84000
26042         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
26044 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
26046         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
26047         to merge probabilities.
26048         * predict.c (probably_never_executed): Also mark as cold functions
26049         with global 0 profile and guessed local profile.
26050         * profile-count.c (profile_probability::combine_with_count): New
26051         member function.
26052         * profile-count.h (profile_probability::operator*,
26053         profile_probability::operator*=, profile_probability::operator/,
26054         profile_probability::operator/=): Reduce precision to adjusted
26055         and set value to guessed on contradictory divisions.
26056         (profile_probability::combine_with_freq): Remove.
26057         (profile_probability::combine_wiht_count): Declare.
26058         (profile_count::force_nonzero):: Set to adjusted.
26059         (profile_count::probability_in):: Set quality to adjusted.
26060         * tree-ssa-tail-merge.c (replace_block_by): Use
26061         combine_with_count.
26063 2018-01-23  Andrew Waterman  <andrew@sifive.com>
26064             Jim Wilson  <jimw@sifive.com>
26066         * config/riscv/riscv.c (riscv_stack_boundary): New.
26067         (riscv_option_override): Set riscv_stack_boundary.  Handle
26068         riscv_preferred_stack_boundary_arg.
26069         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
26070         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
26071         (STACK_BOUNDARY): Set to riscv_stack_boundary.
26072         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
26073         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
26074         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
26076 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
26078         PR target/83905
26079         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
26080         of struct ix86_frame.
26081         (ix86_expand_epilogue): Likewise.  Add a local variable for
26082         the reg_save_offset field in struct ix86_frame.
26084 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
26086         PR tree-optimization/82604
26087         * tree-loop-distribution.c (enum partition_kind): New enum item
26088         PKIND_PARTIAL_MEMSET.
26089         (partition_builtin_p): Support above new enum item.
26090         (generate_code_for_partition): Ditto.
26091         (compute_access_range): Differentiate cases that equality can be
26092         proven at all loops, the innermost loops or no loops.
26093         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
26094         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
26095         (finalize_partitions, distribute_loop): Don't fuse partition of
26096         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
26097         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
26098         parloop is enabled.
26100 2018-01-23  Martin Liska  <mliska@suse.cz>
26102         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
26103         order to ignore the predictor.
26104         (PRED_POLYMORPHIC_CALL): Likewise.
26105         (PRED_RECURSIVE_CALL): Likewise.
26107 2018-01-23  Martin Liska  <mliska@suse.cz>
26109         * tree-profile.c (tree_profiling): Print function header to
26110         aware reader which function we are working on.
26111         * value-prof.c (gimple_find_values_to_profile): Do not print
26112         not interesting value histograms.
26114 2018-01-23  Martin Liska  <mliska@suse.cz>
26116         * profile-count.h (enum profile_quality): Add
26117         profile_uninitialized as the first value. Do not number values
26118         as they are zero based.
26119         (profile_count::verify): Update sanity check.
26120         (profile_probability::verify): Likewise.
26122 2018-01-23  Nathan Sidwell  <nathan@acm.org>
26124         * doc/invoke.texi (ffor-scope): Deprecate.
26126 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
26128         PR tree-optimization/83510
26129         * domwalk.c (set_all_edges_as_executable): New function.
26130         (dom_walker::dom_walker): Convert bool param
26131         "skip_unreachable_blocks" to enum reachability.  Move setup of
26132         edge flags to set_all_edges_as_executable and only do it when
26133         reachability is REACHABLE_BLOCKS.
26134         * domwalk.h (enum dom_walker::reachability): New enum.
26135         (dom_walker::dom_walker): Convert bool param
26136         "skip_unreachable_blocks" to enum reachability.
26137         (set_all_edges_as_executable): New decl.
26138         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
26139         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
26140         "reachability".
26141         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
26142         but converting true to REACHABLE_BLOCKS.
26143         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
26144         * tree-vrp.c
26145         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
26146         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
26147         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
26148         REACHABLE_BLOCKS.
26149         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
26150         if check_all_array_refs will be called.
26152 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
26154         * tree.c (selftest::test_location_wrappers): Add more test
26155         coverage.
26157 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
26159         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
26160         (selftest::test_bit_in_range): Likewise.
26162 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
26164         PR testsuite/83888
26165         * doc/sourcebuild.texi (vect_float): Say that the selector
26166         only describes the situation when -funsafe-math-optimizations is on.
26167         (vect_float_strict): Document.
26169 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
26171         PR tree-optimization/83965
26172         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
26173         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
26174         instead of checking only for a reduction.
26175         (vect_recog_widen_sum_pattern): Likewise.
26177 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
26179         * predict.c (probably_never_executed): Only use precise profile info.
26180         (compute_function_frequency): Skip after inlining hack since we now
26181         have quality checking.
26183 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
26185         * profile-count.h (profile_probability::very_unlikely,
26186         profile_probability::unlikely, profile_probability::even): Set
26187         precision to guessed.
26189 2018-01-23  Richard Biener  <rguenther@suse.de>
26191         PR tree-optimization/83963
26192         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
26193         Properly terminate dominator walk when crossing the exit edge not
26194         when visiting its source block.
26196 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
26198         PR c++/83918
26199         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
26200         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
26202 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
26204         PR tree-optimization/83957
26205         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
26206         semicolon after for body surrounded by braces.
26208         PR tree-optimization/83081
26209         * profile-count.h (profile_probability::split): New method.
26210         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
26211         Use profile_probability::split.
26212         (do_compare_rtx_and_jump): Fix adjustment of probabilities
26213         when splitting a single conditional jump into 2.
26215 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
26217         PR tree-optimization/69452
26218         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
26219         decl.
26221 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26223         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
26224         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
26225         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
26227 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26229         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
26230         declaration.
26231         * config/rl78/rl78.md (movdi): New define_expand.
26232         * config/rl78/rl78.c (rl78_split_movdi): New function.
26234 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
26236         PR target/83862
26237         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
26238         no longer used.
26239         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
26240         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
26241         128-bit to produce an UNSPEC move to get the double word with the
26242         signbit and then a shift directly to do signbit.
26243         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
26244         implementation with a new version that just does either a direct
26245         move or a regular move.  Move memory interface to separate insns.
26246         Move insns so they are next to the expander.
26247         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
26248         with signbit move.  Split big and little endian case.
26249         (signbit<mode>2_dm_mem_le): Likewise.
26250         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
26251         (signbit<mode>2_dm2): Likewise.
26253 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26255         * config/rl78/rl78.md (anddi3): New define_expand.
26257 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26259         * config/rl78/rl78.md (umindi3): New define_expand.
26261 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26263         * config/rl78/rl78.md (smindi3): New define_expand.
26265 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26267         * config/rl78/rl78.md (smaxdi3): New define_expand.
26269 2018-01-22  Carl Love  <cel@us.ibm.com>
26271         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
26272         LVX_V1TI): Add macro expansion.
26273         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
26274         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
26275         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
26276         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
26277         Change check to determine if the instruction is a byte reversing
26278         entry.  Fix typo in comment.
26279         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
26280         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
26281         Add def_builtin calls for new builtins.
26282         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
26283         Add define_insn expansion.
26285 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26287         * config/rl78/rl78.md (umaxdi3): New define_expand.
26289 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
26291         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
26292         for non-QImode registers.
26294 2018-01-22  Richard Biener  <rguenther@suse.de>
26296         PR tree-optimization/83963
26297         * graphite-scop-detection.c (scop_detection::get_sese): Delay
26298         including the loop exit block.
26299         (scop_detection::merge_sese): Likewise.
26300         (scop_detection::add_scop): Do it here instead.
26302 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26304         * doc/sourcebuild.texi (arm_softfloat): Document.
26306 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
26308         PR gcc/77734
26309         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
26310         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
26311         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
26313 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26314             David Edelsohn  <dje.gcc@gmail.com>
26316         PR target/83946
26317         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
26318         Change "crset eq" to "crset 2".
26319         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
26320         (*call_indirect_aix<mode>_nospec): Likewise.
26321         (*call_value_indirect_aix<mode>_nospec): Likewise.
26322         (*call_indirect_elfv2<mode>_nospec): Likewise.
26323         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
26324         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
26325         change assembly output from . to $.
26326         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
26327         (indirect_jump<mode>_nospec): Change assembly output from . to $.
26328         (*tablejump<mode>_internal1_nospec): Likewise.
26330 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
26332         PR target/80870
26333         * config/sh/sh_optimize_sett_clrt.cc:
26334         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
26336 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
26338         PR tree-optimization/83940
26339         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
26340         offset_dt to vect_constant_def rather than vect_unknown_def_type.
26341         (vect_check_load_store_mask): Add a mask_dt_out parameter and
26342         use it to pass back the definition type.
26343         (vect_check_store_rhs): Likewise rhs_dt_out.
26344         (vect_build_gather_load_calls): Add a mask_dt argument and use
26345         it instead of a call to vect_is_simple_use.
26346         (vectorizable_store): Update calls to vect_check_load_store_mask
26347         and vect_check_store_rhs.  Use the dt returned by the latter instead
26348         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
26349         instead of calls to vect_is_simple_use.  Pass the scalar rather
26350         than the vector operand to vect_is_simple_use when handling
26351         second and subsequent copies of an rhs value.
26352         (vectorizable_load): Update calls to vect_check_load_store_mask
26353         and vect_build_gather_load_calls.  Use the cached mask_dt and
26354         gs_info.offset_dt instead of calls to vect_is_simple_use.
26356 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
26358         PR middle-end/83945
26359         * tree-emutls.c: Include gimplify.h.
26360         (lower_emutls_2): New function.
26361         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
26362         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
26363         it before further processing.
26365         PR target/83930
26366         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
26367         UINTVAL (trueop1) instead of INTVAL (op1).
26369 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
26371         PR debug/81570
26372         PR debug/83728
26373         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
26374         INCOMING_FRAME_SP_OFFSET if not defined.
26375         (scan_trace): Add ENTRY argument.  If true and
26376         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
26377         emit a note to adjust the CFA offset.
26378         (create_cfi_notes): Adjust scan_trace callers.
26379         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
26380         INCOMING_FRAME_SP_OFFSET in the CIE.
26381         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
26382         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
26383         Likewise.
26384         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
26385         * doc/tm.texi: Regenerated.
26387 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26389         PR rtl-optimization/83147
26390         * lra-constraints.c (remove_inheritance_pseudos): Use
26391         lra_substitute_pseudo_within_insn.
26393 2018-01-19  Tom de Vries  <tom@codesourcery.com>
26394             Cesar Philippidis  <cesar@codesourcery.com>
26396         PR target/83920
26397         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
26399 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
26401         PR target/83790
26402         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
26403         spaces for function labels.
26405 2018-01-19  Martin Liska  <mliska@suse.cz>
26407         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
26408         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
26409         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
26410         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
26411         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
26412         (PRED_CONST_RETURN): Change from 69 to 65.
26413         (PRED_NULL_RETURN): Change from 91 to 71.
26414         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
26415         (PRED_LOOP_GUARD): Change from 66 to 73.
26417 2018-01-19  Martin Liska  <mliska@suse.cz>
26419         * predict.c (predict_insn_def): Add new assert.
26420         (struct branch_predictor): Change type to signed integer.
26421         (test_prediction_value_range): Amend test to cover
26422         PROB_UNINITIALIZED.
26423         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
26424         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
26425         (PRED_LOOP_ITERATIONS_MAX): Likewise.
26426         (PRED_LOOP_IV_COMPARE): Likewise.
26427         * predict.h (PROB_UNINITIALIZED): Define new constant.
26429 2018-01-19  Martin Liska  <mliska@suse.cz>
26431         * predict.c (dump_prediction): Add new format for
26432         analyze_brprob.py script which is enabled with -details
26433         suboption.
26434         * profile-count.h (precise_p): New function.
26436 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
26438         PR tree-optimization/83922
26439         * tree-vect-loop.c (vect_verify_full_masking): Return false if
26440         there are no statements that need masking.
26441         (vect_active_double_reduction_p): New function.
26442         (vect_analyze_loop_operations): Use it when handling phis that
26443         are not in the loop header.
26445 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
26447         PR tree-optimization/83914
26448         * tree-vect-loop.c (vectorizable_induction): Don't convert
26449         init_expr or apply the peeling adjustment for inductions
26450         that are nested within the vectorized loop.
26452 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26454         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
26455         instead of NEG.
26457 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
26459         PR sanitizer/81715
26460         PR testsuite/83882
26461         * function.h (gimplify_parameters): Add gimple_seq * argument.
26462         * function.c: Include gimple.h and options.h.
26463         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
26464         for the added local temporaries if needed.
26465         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
26466         if there are any parameter cleanups, wrap whole body into a
26467         try/finally with the cleanups.
26469 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
26471         PR target/82964
26472         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
26473         Use GET_MODE_CLASS for scalar floating point.
26475 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
26477         PR ipa/82256
26478         patch by PaX Team
26479         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
26480         Fix call of call_cgraph_insertion_hooks.
26482 2018-01-18  Martin Sebor  <msebor@redhat.com>
26484         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
26486 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
26488         PR ipa/83619
26489         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
26490         frequencies.
26492 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
26494         PR other/70268
26495         * common.opt: (-ffile-prefix-map): New option.
26496         * opts.c (common_handle_option): Defer it.
26497         * opts-global.c (handle_common_deferred_options): Handle it.
26498         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
26499         * file-prefix-map.h: New file.
26500         (remap_debug_filename, add_debug_prefix_map): ...here.
26501         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
26502         * final.c (debug_prefix_map, add_debug_prefix_map
26503         remap_debug_filename): Move to...
26504         * file-prefix-map.c: New file.
26505         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
26506         generalize, get rid of alloca(), use strrchr() instead of strchr().
26507         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
26508         Implement in terms of add_prefix_map().
26509         (remap_macro_filename, remap_debug_filename): Implement in term of
26510         remap_filename().
26511         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
26512         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
26513         * dbxout.c: Include file-prefix-map.h.
26514         * varasm.c: Likewise.
26515         * vmsdbgout.c: Likewise.
26516         * xcoffout.c: Likewise.
26517         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
26518         * doc/cppopts.texi (-fmacro-prefix-map): Document.
26519         * doc/invoke.texi (-ffile-prefix-map): Document.
26520         (-fdebug-prefix-map): Update description.
26522 2018-01-18  Martin Liska  <mliska@suse.cz>
26524         * config/i386/i386.c (indirect_thunk_name): Document that also
26525         lfence is emitted.
26526         (output_indirect_thunk): Document why both instructions
26527         (pause and lfence) are generated.
26529 2018-01-18  Richard Biener  <rguenther@suse.de>
26531         PR tree-optimization/83887
26532         * graphite-scop-detection.c
26533         (scop_detection::get_nearest_dom_with_single_entry): Remove.
26534         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
26535         (scop_detection::merge_sese): Re-implement with a flood-fill
26536         algorithm that properly finds a SESE region if it exists.
26538 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
26540         PR c/61240
26541         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
26542         pointer_diff optimizations use view_convert instead of convert.
26544 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26546         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
26547         Generate different code for -mno-speculate-indirect-jumps.
26548         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
26549         (*call_indirect_aix<mode>): Disable for
26550         -mno-speculate-indirect-jumps.
26551         (*call_indirect_aix<mode>_nospec): New define_insn.
26552         (*call_value_indirect_aix<mode>): Disable for
26553         -mno-speculate-indirect-jumps.
26554         (*call_value_indirect_aix<mode>_nospec): New define_insn.
26555         (*sibcall_nonlocal_sysv<mode>): Generate different code for
26556         -mno-speculate-indirect-jumps.
26557         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
26559 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
26561         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
26562         long double type, set the flags for noting the default long double
26563         type, even if we don't pass or return a long double type.
26565 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
26567         PR ipa/83051
26568         * ipa-inline.c (flatten_function): Do not overwrite final inlining
26569         failure.
26571 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
26573         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
26574         support for merge[hl].
26575         (fold_mergehl_helper): New helper function.
26576         (tree-vector-builder.h): New #include for tree_vector_builder usage.
26577         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
26578         (altivec_vmrglw_direct): Add xxmrglw insn.
26580 2018-01-17  Andrew Waterman  <andrew@sifive.com>
26582         * config/riscv/riscv.c (riscv_conditional_register_usage): If
26583         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
26585 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
26587         PR lto/83121
26588         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
26589         call the lto_location_cache before reading the
26590         DECL_SOURCE_LOCATION of the types.
26592 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
26593             Richard Sandiford  <richard.sandiford@linaro.org>
26595         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
26596         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
26597         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
26598         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
26599         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
26600         Add declaration.
26601         * config/aarch64/constraints.md (aarch64_movti_operand):
26602         Limit immediates.
26603         * config/aarch64/predicates.md (Uti): Add new constraint.
26605 2018-01-17  Carl Love  <cel@us.ibm.com>
26607         * config/rs6000/vsx.md (define_expand xl_len_r,
26608         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
26609         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
26610         lxvll.
26611         (define_expand, define_insn): Move the shift left from  the
26612         define_insn to the define_expand for lxvl and stxvl instructions.
26613         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
26614         and XL_LEN_R definitions to PURE.
26616 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
26618         * config/i386/i386.c (indirect_thunk_name): Declare regno
26619         as unsigned int.  Compare regno with INVALID_REGNUM.
26620         (output_indirect_thunk): Ditto.
26621         (output_indirect_thunk_function): Ditto.
26622         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
26623         in the call to output_indirect_thunk_function.
26625 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
26627         PR middle-end/83884
26628         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
26629         rather than the size of inner_type to determine the stack slot size
26630         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
26632 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
26634         PR target/83546
26635         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
26636         to PTA_SILVERMONT.
26638 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
26640         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
26641         endian Linux systems to optionally enable multilibs for selecting
26642         the long double type if the user configured an explicit type.
26643         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
26644         have no long double multilibs if not defined.
26645         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
26646         warn if the user used -mabi={ieee,ibm}longdouble and we built
26647         multilibs for long double.
26648         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
26649         appropriate multilib option.
26650         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
26651         multilib options.
26652         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
26653         for building long double multilibs.
26654         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
26656 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
26658         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
26659         copies.
26661         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
26662         64 bits.
26663         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
26664         128 bits.
26666         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
26667         variables.
26669         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
26670         return value.
26672 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
26674         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
26675         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
26677 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
26679         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
26680         different rtl trees depending on TARGET_64BIT.
26681         (rs6000_gen_lvx): Likewise.
26683 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
26685         * config/visium/visium.md (nop): Tweak comment.
26686         (hazard_nop): Likewise.
26688 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26690         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
26691         -mspeculate-indirect-jumps.
26692         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
26693         for -mno-speculate-indirect-jumps.
26694         (*call_indirect_elfv2<mode>_nospec): New define_insn.
26695         (*call_value_indirect_elfv2<mode>): Disable for
26696         -mno-speculate-indirect-jumps.
26697         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
26698         (indirect_jump): Emit different RTL for
26699         -mno-speculate-indirect-jumps.
26700         (*indirect_jump<mode>): Disable for
26701         -mno-speculate-indirect-jumps.
26702         (*indirect_jump<mode>_nospec): New define_insn.
26703         (tablejump): Emit different RTL for
26704         -mno-speculate-indirect-jumps.
26705         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
26706         (tablejumpsi_nospec): New define_expand.
26707         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
26708         (tablejumpdi_nospec): New define_expand.
26709         (*tablejump<mode>_internal1): Disable for
26710         -mno-speculate-indirect-jumps.
26711         (*tablejump<mode>_internal1_nospec): New define_insn.
26712         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
26713         option.
26715 2018-01-16  Artyom Skrobov tyomitch@gmail.com
26717         * caller-save.c (insert_save): Drop unnecessary parameter.  All
26718         callers updated.
26720 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
26721             Richard Biener  <rguenth@suse.de>
26723         PR libgomp/83590
26724         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
26725         return early, inline manually is_gimple_sizepos.  Make sure if we
26726         call gimplify_expr we don't end up with a gimple constant.
26727         * tree.c (variably_modified_type_p): Don't return true for
26728         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
26729         * gimplify.h (is_gimple_sizepos): Remove.
26731 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
26733         PR tree-optimization/83857
26734         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
26735         vectorizable_live_operation for pure SLP statements.
26736         (vectorizable_live_operation): Handle PHIs.
26738 2018-01-16  Richard Biener  <rguenther@suse.de>
26740         PR tree-optimization/83867
26741         * tree-vect-stmts.c (vect_transform_stmt): Precompute
26742         nested_in_vect_loop_p since the scalar stmt may get invalidated.
26744 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
26746         PR c/83844
26747         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
26748         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
26749         If off is not INTEGER_CST, issue a may not be aligned warning
26750         rather than isn't aligned.  Use isn%'t rather than isn't.
26751         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
26752         into MULT_EXPR.
26753         <case MULT_EXPR>: Improve the case when bottom and one of the
26754         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
26755         operand, in that case check if the other operand is multiple of
26756         bottom divided by the INTEGER_CST operand.
26758 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
26760         PR target/83858
26761         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
26762         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
26763         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
26764         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
26765         * config/pa/pa.c (pa_function_arg_advance): Likewise.
26766         (pa_function_arg, pa_arg_partial_bytes): Likewise.
26767         (pa_function_arg_size): New function.
26769 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
26771         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
26772         in a separate statement.
26774 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
26776         PR tree-optimization/83847
26777         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
26778         group gathers and scatters.
26780 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
26782         PR rtl-optimization/86620
26783         * params.def (max-sched-ready-insns): Bump minimum value to 1.
26785         PR rtl-optimization/83213
26786         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
26787         to last if both are JUMP_INSNs.
26789         PR tree-optimization/83843
26790         * gimple-ssa-store-merging.c
26791         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
26792         store_immediate_info for bswap/nop orig_stores.
26794 2018-01-15  Andrew Waterman  <andrew@sifive.com>
26796         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
26797         !TARGET_MUL.
26798         <UDIV>: Increase cost if !TARGET_DIV.
26800 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
26802         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
26803         (define_attr "cr_logical_3op"): New.
26804         (cceq_ior_compare): Adjust.
26805         (cceq_ior_compare_complement): Adjust.
26806         (*cceq_rev_compare): Adjust.
26807         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
26808         (is_cracked_insn): Adjust.
26809         (insn_must_be_first_in_group): Adjust.
26810         * config/rs6000/40x.md: Adjust.
26811         * config/rs6000/440.md: Adjust.
26812         * config/rs6000/476.md: Adjust.
26813         * config/rs6000/601.md: Adjust.
26814         * config/rs6000/603.md: Adjust.
26815         * config/rs6000/6xx.md: Adjust.
26816         * config/rs6000/7450.md: Adjust.
26817         * config/rs6000/7xx.md: Adjust.
26818         * config/rs6000/8540.md: Adjust.
26819         * config/rs6000/cell.md: Adjust.
26820         * config/rs6000/e300c2c3.md: Adjust.
26821         * config/rs6000/e500mc.md: Adjust.
26822         * config/rs6000/e500mc64.md: Adjust.
26823         * config/rs6000/e5500.md: Adjust.
26824         * config/rs6000/e6500.md: Adjust.
26825         * config/rs6000/mpc.md: Adjust.
26826         * config/rs6000/power4.md: Adjust.
26827         * config/rs6000/power5.md: Adjust.
26828         * config/rs6000/power6.md: Adjust.
26829         * config/rs6000/power7.md: Adjust.
26830         * config/rs6000/power8.md: Adjust.
26831         * config/rs6000/power9.md: Adjust.
26832         * config/rs6000/rs64.md: Adjust.
26833         * config/rs6000/titan.md: Adjust.
26835 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
26837         * config/i386/predicates.md (indirect_branch_operand): Rewrite
26838         ix86_indirect_branch_register logic.
26840 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
26842         * config/i386/constraints.md (Bs): Update
26843         ix86_indirect_branch_register check.  Don't check
26844         ix86_indirect_branch_register with GOT_memory_operand.
26845         (Bw): Likewise.
26846         * config/i386/predicates.md (GOT_memory_operand): Don't check
26847         ix86_indirect_branch_register here.
26848         (GOT32_symbol_operand): Likewise.
26850 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
26852         * config/i386/predicates.md (constant_call_address_operand):
26853         Rewrite ix86_indirect_branch_register logic.
26854         (sibcall_insn_operand): Likewise.
26856 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
26858         * config/i386/constraints.md (Bs): Replace
26859         ix86_indirect_branch_thunk_register with
26860         ix86_indirect_branch_register.
26861         (Bw): Likewise.
26862         * config/i386/i386.md (indirect_jump): Likewise.
26863         (tablejump): Likewise.
26864         (*sibcall_memory): Likewise.
26865         (*sibcall_value_memory): Likewise.
26866         Peepholes of indirect call and jump via memory: Likewise.
26867         * config/i386/i386.opt: Likewise.
26868         * config/i386/predicates.md (indirect_branch_operand): Likewise.
26869         (GOT_memory_operand): Likewise.
26870         (call_insn_operand): Likewise.
26871         (sibcall_insn_operand): Likewise.
26872         (GOT32_symbol_operand): Likewise.
26874 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
26876         PR middle-end/83837
26877         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
26878         type rather than type addr's type points to.
26879         (expand_omp_atomic_mutex): Likewise.
26880         (expand_omp_atomic): Likewise.
26882 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
26884         PR target/83839
26885         * config/i386/i386.c (output_indirect_thunk_function): Use
26886         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
26887         for  __x86_return_thunk.
26889 2018-01-15  Richard Biener  <rguenther@suse.de>
26891         PR middle-end/83850
26892         * expmed.c (extract_bit_field_1): Fix typo.
26894 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26896         PR target/83687
26897         * config/arm/iterators.md (VF): New mode iterator.
26898         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
26899         Remove integer-related logic from pattern.
26900         (neon_vabd<mode>_3): Likewise.
26902 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
26904         PR middle-end/82694
26905         * common.opt (fstrict-overflow): No longer an alias.
26906         (fwrapv-pointer): New option.
26907         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
26908         also for pointer types based on flag_wrapv_pointer.
26909         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
26910         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
26911         opts->x_flag_wrapv got set.
26912         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
26913         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
26914         POINTER_TYPE_OVERFLOW_UNDEFINED.
26915         * match.pd: Likewise in address comparison pattern.
26916         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
26918 2018-01-15  Richard Biener  <rguenther@suse.de>
26920         PR lto/83804
26921         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
26922         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
26923         Reset type names to their identifier if their TYPE_DECL doesn't
26924         have linkage (and thus is used for ODR and devirt).
26925         (save_debug_info_for_decl): Remove.
26926         (save_debug_info_for_type): Likewise.
26927         (add_tree_to_fld_list): Adjust.
26928         * tree-pretty-print.c (dump_generic_node): Make dumping of
26929         type names more robust.
26931 2018-01-15  Richard Biener  <rguenther@suse.de>
26933         * BASE-VER: Bump to 8.0.1.
26935 2018-01-14  Martin Sebor  <msebor@redhat.com>
26937         PR other/83508
26938         * builtins.c (check_access): Avoid warning when the no-warning bit
26939         is set.
26941 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
26943         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
26944         * ira-color (allocno_hard_regs_compare): Likewise.
26946 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
26948         PR target/83013
26949         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
26950         Use .pushsection/.popsection.
26952 2018-01-14  Martin Sebor  <msebor@redhat.com>
26954         PR c++/81327
26955         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
26957 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
26959         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
26960         entry from extra_headers.
26961         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
26962         extra_headers, make the list bitwise identical to the i?86-*-* one.
26964 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
26966         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
26967         -mcmodel=large with -mindirect-branch=thunk,
26968         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
26969         -mfunction-return=thunk-extern.
26970         * doc/invoke.texi: Document -mcmodel=large is incompatible with
26971         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
26972         -mfunction-return=thunk and -mfunction-return=thunk-extern.
26974 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
26976         * config/i386/i386.c (print_reg): Print the name of the full
26977         integer register without '%'.
26978         (ix86_print_operand): Handle 'V'.
26979         * doc/extend.texi: Document 'V' modifier.
26981 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
26983         * config/i386/constraints.md (Bs): Disallow memory operand for
26984         -mindirect-branch-register.
26985         (Bw): Likewise.
26986         * config/i386/predicates.md (indirect_branch_operand): Likewise.
26987         (GOT_memory_operand): Likewise.
26988         (call_insn_operand): Likewise.
26989         (sibcall_insn_operand): Likewise.
26990         (GOT32_symbol_operand): Likewise.
26991         * config/i386/i386.md (indirect_jump): Call convert_memory_address
26992         for -mindirect-branch-register.
26993         (tablejump): Likewise.
26994         (*sibcall_memory): Likewise.
26995         (*sibcall_value_memory): Likewise.
26996         Disallow peepholes of indirect call and jump via memory for
26997         -mindirect-branch-register.
26998         (*call_pop): Replace m with Bw.
26999         (*call_value_pop): Likewise.
27000         (*sibcall_pop_memory): Replace m with Bs.
27001         * config/i386/i386.opt (mindirect-branch-register): New option.
27002         * doc/invoke.texi: Document -mindirect-branch-register option.
27004 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27006         * config/i386/i386-protos.h (ix86_output_function_return): New.
27007         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
27008         set function_return_type.
27009         (indirect_thunk_name): Add ret_p to indicate thunk for function
27010         return.
27011         (output_indirect_thunk_function): Pass false to
27012         indirect_thunk_name.
27013         (ix86_output_indirect_branch_via_reg): Likewise.
27014         (ix86_output_indirect_branch_via_push): Likewise.
27015         (output_indirect_thunk_function): Create alias for function
27016         return thunk if regno < 0.
27017         (ix86_output_function_return): New function.
27018         (ix86_handle_fndecl_attribute): Handle function_return.
27019         (ix86_attribute_table): Add function_return.
27020         * config/i386/i386.h (machine_function): Add
27021         function_return_type.
27022         * config/i386/i386.md (simple_return_internal): Use
27023         ix86_output_function_return.
27024         (simple_return_internal_long): Likewise.
27025         * config/i386/i386.opt (mfunction-return=): New option.
27026         (indirect_branch): Mention -mfunction-return=.
27027         * doc/extend.texi: Document function_return function attribute.
27028         * doc/invoke.texi: Document -mfunction-return= option.
27030 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
27032         * config/i386/i386-opts.h (indirect_branch): New.
27033         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
27034         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
27035         with local indirect jump when converting indirect call and jump.
27036         (ix86_set_indirect_branch_type): New.
27037         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
27038         (indirectlabelno): New.
27039         (indirect_thunk_needed): Likewise.
27040         (indirect_thunk_bnd_needed): Likewise.
27041         (indirect_thunks_used): Likewise.
27042         (indirect_thunks_bnd_used): Likewise.
27043         (INDIRECT_LABEL): Likewise.
27044         (indirect_thunk_name): Likewise.
27045         (output_indirect_thunk): Likewise.
27046         (output_indirect_thunk_function): Likewise.
27047         (ix86_output_indirect_branch_via_reg): Likewise.
27048         (ix86_output_indirect_branch_via_push): Likewise.
27049         (ix86_output_indirect_branch): Likewise.
27050         (ix86_output_indirect_jmp): Likewise.
27051         (ix86_code_end): Call output_indirect_thunk_function if needed.
27052         (ix86_output_call_insn): Call ix86_output_indirect_branch if
27053         needed.
27054         (ix86_handle_fndecl_attribute): Handle indirect_branch.
27055         (ix86_attribute_table): Add indirect_branch.
27056         * config/i386/i386.h (machine_function): Add indirect_branch_type
27057         and has_local_indirect_jump.
27058         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
27059         to true.
27060         (tablejump): Likewise.
27061         (*indirect_jump): Use ix86_output_indirect_jmp.
27062         (*tablejump_1): Likewise.
27063         (simple_return_indirect_internal): Likewise.
27064         * config/i386/i386.opt (mindirect-branch=): New option.
27065         (indirect_branch): New.
27066         (keep): Likewise.
27067         (thunk): Likewise.
27068         (thunk-inline): Likewise.
27069         (thunk-extern): Likewise.
27070         * doc/extend.texi: Document indirect_branch function attribute.
27071         * doc/invoke.texi: Document -mindirect-branch= option.
27073 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
27075         PR ipa/83051
27076         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
27078 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
27080         * ipa-inline.c (want_inline_small_function_p): Return false if
27081         inlining has already failed with CIF_FINAL_ERROR.
27082         (update_caller_keys): Call want_inline_small_function_p before
27083         can_inline_edge_p.
27084         (update_callee_keys): Likewise.
27086 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27088         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
27089         New function.
27090         (rs6000_quadword_masked_address_p): Likewise.
27091         (quad_aligned_load_p): Likewise.
27092         (quad_aligned_store_p): Likewise.
27093         (const_load_sequence_p): Add comment to describe the outer-most loop.
27094         (mimic_memory_attributes_and_flags): New function.
27095         (rs6000_gen_stvx): Likewise.
27096         (replace_swapped_aligned_store): Likewise.
27097         (rs6000_gen_lvx): Likewise.
27098         (replace_swapped_aligned_load): Likewise.
27099         (replace_swapped_load_constant): Capitalize argument name in
27100         comment describing this function.
27101         (rs6000_analyze_swaps): Add a third pass to search for vector loads
27102         and stores that access quad-word aligned addresses and replace
27103         with stvx or lvx instructions when appropriate.
27104         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
27105         New function prototype.
27106         (rs6000_quadword_masked_address_p): Likewise.
27107         (rs6000_gen_lvx): Likewise.
27108         (rs6000_gen_stvx): Likewise.
27109         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
27110         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
27111         when memory address is aligned.
27112         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
27113         this split to select lvx instruction when memory address is aligned.
27114         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
27115         instruction when memory address is aligned.
27116         (*vsx_le_perm_load_v16qi): Likewise.
27117         (four unnamed splitters): Modify to select the stvx instruction
27118         when memory is aligned.
27120 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
27122         * predict.c (determine_unlikely_bbs): Handle correctly BBs
27123         which appears in the queue multiple times.
27125 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27126             Alan Hayward  <alan.hayward@arm.com>
27127             David Sherwood  <david.sherwood@arm.com>
27129         * tree-vectorizer.h (vec_lower_bound): New structure.
27130         (_loop_vec_info): Add check_nonzero and lower_bounds.
27131         (LOOP_VINFO_CHECK_NONZERO): New macro.
27132         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
27133         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
27134         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
27135         fields.  Make seg_len the distance travelled, not including the
27136         access size.
27137         (dr_direction_indicator): Declare.
27138         (dr_zero_step_indicator): Likewise.
27139         (dr_known_forward_stride_p): Likewise.
27140         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
27141         tree-ssanames.h.
27142         (runtime_alias_check_p): Allow runtime alias checks with
27143         variable strides.
27144         (operator ==): Compare access_size and align.
27145         (prune_runtime_alias_test_list): Rework for new distinction between
27146         the access_size and seg_len.
27147         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
27148         segment lengths.
27149         (get_segment_min_max): New function.
27150         (create_intersect_range_checks): Use it.
27151         (dr_step_indicator): New function.
27152         (dr_direction_indicator): Likewise.
27153         (dr_zero_step_indicator): Likewise.
27154         (dr_known_forward_stride_p): Likewise.
27155         * tree-loop-distribution.c (data_ref_segment_size): Return
27156         DR_STEP * (niters - 1).
27157         (compute_alias_check_pairs): Update call to the dr_with_seg_len
27158         constructor.
27159         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
27160         (vect_preserves_scalar_order_p): New function, split out from...
27161         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
27162         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
27163         (vect_vfa_access_size): New function.
27164         (vect_vfa_align): Likewise.
27165         (vect_compile_time_alias): Take access_size_a and access_b arguments.
27166         (dump_lower_bound): New function.
27167         (vect_check_lower_bound): Likewise.
27168         (vect_small_gap_p): Likewise.
27169         (vectorizable_with_step_bound_p): Likewise.
27170         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
27171         depencies if the vectorization factor is 1.  Convert the checks
27172         for nonzero steps into checks on the bounds of DR_STEP.  Try using
27173         a bunds check for variable steps if the minimum required step is
27174         relatively small. Update calls to the dr_with_seg_len
27175         constructor and to vect_compile_time_alias.
27176         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
27177         function.
27178         (vect_loop_versioning): Call it.
27179         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
27180         when retrying.
27181         (vect_estimate_min_profitable_iters): Account for any bounds checks.
27183 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27184             Alan Hayward  <alan.hayward@arm.com>
27185             David Sherwood  <david.sherwood@arm.com>
27187         * doc/sourcebuild.texi (vect_scatter_store): Document.
27188         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
27189         optabs.
27190         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
27191         Document.
27192         * genopinit.c (main): Add supports_vec_scatter_store and
27193         supports_vec_scatter_store_cached to target_optabs.
27194         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
27195         IFN_MASK_SCATTER_STORE.
27196         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
27197         functions.
27198         * internal-fn.h (internal_store_fn_p): Declare.
27199         (internal_fn_stored_value_index): Likewise.
27200         * internal-fn.c (scatter_store_direct): New macro.
27201         (expand_scatter_store_optab_fn): New function.
27202         (direct_scatter_store_optab_supported_p): New macro.
27203         (internal_store_fn_p): New function.
27204         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
27205         IFN_MASK_SCATTER_STORE.
27206         (internal_fn_mask_index): Likewise.
27207         (internal_fn_stored_value_index): New function.
27208         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
27209         for scatter stores.
27210         * optabs-query.h (supports_vec_scatter_store_p): Declare.
27211         * optabs-query.c (supports_vec_scatter_store_p): New function.
27212         * tree-vectorizer.h (vect_get_store_rhs): Declare.
27213         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
27214         true for scatter stores.
27215         (vect_gather_scatter_fn_p): Handle scatter stores too.
27216         (vect_check_gather_scatter): Consider using scatter stores if
27217         supports_vec_scatter_store_p.
27218         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
27219         scatter stores too.
27220         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
27221         internal_fn_stored_value_index.
27222         (check_load_store_masking): Handle scatter stores too.
27223         (vect_get_store_rhs): Make public.
27224         (vectorizable_call): Use internal_store_fn_p.
27225         (vectorizable_store): Handle scatter store internal functions.
27226         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
27227         when deciding whether the end of the group has been reached.
27228         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
27229         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
27230         (mask_scatter_store<mode>): New insns.
27232 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27233             Alan Hayward  <alan.hayward@arm.com>
27234             David Sherwood  <david.sherwood@arm.com>
27236         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
27237         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
27238         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
27239         function.
27240         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
27241         Use vect_truncate_gather_scatter_offset if we can't treat the
27242         operation as a normal gather load or scatter store.
27243         (get_group_load_store_type): Take the gather_scatter_info
27244         as argument.  Try using a gather load or scatter store for
27245         single-element groups.
27246         (get_load_store_type): Update calls to get_group_load_store_type
27247         and vect_use_strided_gather_scatters_p.
27249 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27250             Alan Hayward  <alan.hayward@arm.com>
27251             David Sherwood  <david.sherwood@arm.com>
27253         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
27254         optional tree argument.
27255         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
27256         null target hooks.
27257         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
27258         but continue to use the current value as a fallback.
27259         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
27260         to compare the updates.
27261         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
27262         (get_load_store_type): Use it when handling a strided access.
27263         (vect_get_strided_load_store_ops): New function.
27264         (vect_get_data_ptr_increment): Likewise.
27265         (vectorizable_load): Handle strided gather loads.  Always pass
27266         a step to vect_create_data_ref_ptr and bump_vector_ptr.
27268 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27269             Alan Hayward  <alan.hayward@arm.com>
27270             David Sherwood  <david.sherwood@arm.com>
27272         * doc/md.texi (gather_load@var{m}): Document.
27273         (mask_gather_load@var{m}): Likewise.
27274         * genopinit.c (main): Add supports_vec_gather_load and
27275         supports_vec_gather_load_cached to target_optabs.
27276         * optabs-tree.c (init_tree_optimization_optabs): Use
27277         ggc_cleared_alloc to allocate target_optabs.
27278         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
27279         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
27280         functions.
27281         * internal-fn.h (internal_load_fn_p): Declare.
27282         (internal_gather_scatter_fn_p): Likewise.
27283         (internal_fn_mask_index): Likewise.
27284         (internal_gather_scatter_fn_supported_p): Likewise.
27285         * internal-fn.c (gather_load_direct): New macro.
27286         (expand_gather_load_optab_fn): New function.
27287         (direct_gather_load_optab_supported_p): New macro.
27288         (direct_internal_fn_optab): New function.
27289         (internal_load_fn_p): Likewise.
27290         (internal_gather_scatter_fn_p): Likewise.
27291         (internal_fn_mask_index): Likewise.
27292         (internal_gather_scatter_fn_supported_p): Likewise.
27293         * optabs-query.c (supports_at_least_one_mode_p): New function.
27294         (supports_vec_gather_load_p): Likewise.
27295         * optabs-query.h (supports_vec_gather_load_p): Declare.
27296         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
27297         and memory_type field.
27298         (NUM_PATTERNS): Bump to 15.
27299         * tree-vect-data-refs.c: Include internal-fn.h.
27300         (vect_gather_scatter_fn_p): New function.
27301         (vect_describe_gather_scatter_call): Likewise.
27302         (vect_check_gather_scatter): Try using internal functions for
27303         gather loads.  Recognize existing calls to a gather load function.
27304         (vect_analyze_data_refs): Consider using gather loads if
27305         supports_vec_gather_load_p.
27306         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
27307         (vect_get_gather_scatter_offset_type): Likewise.
27308         (vect_convert_mask_for_vectype): Likewise.
27309         (vect_add_conversion_to_patterm): Likewise.
27310         (vect_try_gather_scatter_pattern): Likewise.
27311         (vect_recog_gather_scatter_pattern): New pattern recognizer.
27312         (vect_vect_recog_func_ptrs): Add it.
27313         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
27314         internal_fn_mask_index and internal_gather_scatter_fn_p.
27315         (check_load_store_masking): Take the gather_scatter_info as an
27316         argument and handle gather loads.
27317         (vect_get_gather_scatter_ops): New function.
27318         (vectorizable_call): Check internal_load_fn_p.
27319         (vectorizable_load): Likewise.  Handle gather load internal
27320         functions.
27321         (vectorizable_store): Update call to check_load_store_masking.
27322         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
27323         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
27324         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
27325         (aarch64_gather_scale_operand_d): New predicates.
27326         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
27327         (mask_gather_load<mode>): New insns.
27329 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27330             Alan Hayward  <alan.hayward@arm.com>
27331             David Sherwood  <david.sherwood@arm.com>
27333         * optabs.def (fold_left_plus_optab): New optab.
27334         * doc/md.texi (fold_left_plus_@var{m}): Document.
27335         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
27336         * internal-fn.c (fold_left_direct): Define.
27337         (expand_fold_left_optab_fn): Likewise.
27338         (direct_fold_left_optab_supported_p): Likewise.
27339         * fold-const-call.c (fold_const_fold_left): New function.
27340         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
27341         * tree-parloops.c (valid_reduction_p): New function.
27342         (gather_scalar_reductions): Use it.
27343         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
27344         (vect_finish_replace_stmt): Declare.
27345         * tree-vect-loop.c (fold_left_reduction_fn): New function.
27346         (needs_fold_left_reduction_p): New function, split out from...
27347         (vect_is_simple_reduction): ...here.  Accept reductions that
27348         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
27349         (vect_force_simple_reduction): Also store the reduction type in
27350         the assignment's STMT_VINFO_REDUC_TYPE.
27351         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
27352         (merge_with_identity): New function.
27353         (vect_expand_fold_left): Likewise.
27354         (vectorize_fold_left_reduction): Likewise.
27355         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
27356         scalar phi in place for it.  Check for target support and reject
27357         cases that would reassociate the operation.  Defer the transform
27358         phase to vectorize_fold_left_reduction.
27359         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
27360         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
27361         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
27363 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27365         * tree-if-conv.c (predicate_mem_writes): Remove redundant
27366         call to ifc_temp_var.
27368 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27369             Alan Hayward  <alan.hayward@arm.com>
27370             David Sherwood  <david.sherwood@arm.com>
27372         * target.def (legitimize_address_displacement): Take the original
27373         offset as a poly_int.
27374         * targhooks.h (default_legitimize_address_displacement): Update
27375         accordingly.
27376         * targhooks.c (default_legitimize_address_displacement): Likewise.
27377         * doc/tm.texi: Regenerate.
27378         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
27379         as an argument, moving assert of ad->disp == ad->disp_term to...
27380         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
27381         Try calling targetm.legitimize_address_displacement before expanding
27382         the address rather than afterwards, and adjust for the new interface.
27383         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
27384         Match the new hook interface.  Handle SVE addresses.
27385         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
27386         new hook interface.
27388 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27390         * Makefile.in (OBJS): Add early-remat.o.
27391         * target.def (select_early_remat_modes): New hook.
27392         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
27393         * doc/tm.texi: Regenerate.
27394         * targhooks.h (default_select_early_remat_modes): Declare.
27395         * targhooks.c (default_select_early_remat_modes): New function.
27396         * timevar.def (TV_EARLY_REMAT): New timevar.
27397         * passes.def (pass_early_remat): New pass.
27398         * tree-pass.h (make_pass_early_remat): Declare.
27399         * early-remat.c: New file.
27400         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
27401         function.
27402         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
27404 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27405             Alan Hayward  <alan.hayward@arm.com>
27406             David Sherwood  <david.sherwood@arm.com>
27408         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
27409         vfm1 with a bound_epilog parameter.
27410         (vect_do_peeling): Update calls accordingly, and move the prologue
27411         call earlier in the function.  Treat the base bound_epilog as 0 for
27412         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
27413         this base when peeling for gaps.
27414         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
27415         with fully-masked loops.
27416         (vect_estimate_min_profitable_iters): Handle the single peeled
27417         iteration in that case.
27419 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27420             Alan Hayward  <alan.hayward@arm.com>
27421             David Sherwood  <david.sherwood@arm.com>
27423         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
27424         single-element interleaving even if the size is not a power of 2.
27425         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
27426         accesses for single-element interleaving if the group size is
27427         not a power of 2.
27429 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27430             Alan Hayward  <alan.hayward@arm.com>
27431             David Sherwood  <david.sherwood@arm.com>
27433         * doc/md.texi (fold_extract_last_@var{m}): Document.
27434         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
27435         * optabs.def (fold_extract_last_optab): New optab.
27436         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
27437         * internal-fn.c (fold_extract_direct): New macro.
27438         (expand_fold_extract_optab_fn): Likewise.
27439         (direct_fold_extract_optab_supported_p): Likewise.
27440         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
27441         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27442         EXTRACT_LAST_REDUCTION.
27443         (get_initial_def_for_reduction): Do not create an initial vector
27444         for EXTRACT_LAST_REDUCTION reductions.
27445         (vectorizable_reduction): Leave the scalar phi in place for
27446         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
27447         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
27448         epilogue code for EXTRACT_LAST_REDUCTION and defer the
27449         transform phase to vectorizable_condition.
27450         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
27451         split out from...
27452         (vect_finish_stmt_generation): ...here.
27453         (vect_finish_replace_stmt): New function.
27454         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
27455         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
27456         pattern.
27457         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
27459 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27460             Alan Hayward  <alan.hayward@arm.com>
27461             David Sherwood  <david.sherwood@arm.com>
27463         * doc/md.texi (extract_last_@var{m}): Document.
27464         * optabs.def (extract_last_optab): New optab.
27465         * internal-fn.def (EXTRACT_LAST): New internal function.
27466         * internal-fn.c (cond_unary_direct): New macro.
27467         (expand_cond_unary_optab_fn): Likewise.
27468         (direct_cond_unary_optab_supported_p): Likewise.
27469         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
27470         loops using EXTRACT_LAST.
27471         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
27472         (extract_last_<mode>): ...this optab.
27473         (vec_extract<mode><Vel>): Update accordingly.
27475 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27476             Alan Hayward  <alan.hayward@arm.com>
27477             David Sherwood  <david.sherwood@arm.com>
27479         * target.def (empty_mask_is_expensive): New hook.
27480         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
27481         * doc/tm.texi: Regenerate.
27482         * targhooks.h (default_empty_mask_is_expensive): Declare.
27483         * targhooks.c (default_empty_mask_is_expensive): New function.
27484         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
27485         if the target says that empty masks are expensive.
27486         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
27487         New function.
27488         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
27490 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27491             Alan Hayward  <alan.hayward@arm.com>
27492             David Sherwood  <david.sherwood@arm.com>
27494         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
27495         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
27496         (vect_use_loop_mask_for_alignment_p): New function.
27497         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
27498         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
27499         niters_skip argument.  Make sure that the first niters_skip elements
27500         of the first iteration are inactive.
27501         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
27502         Update call to vect_set_loop_masks_directly.
27503         (get_misalign_in_elems): New function, split out from...
27504         (vect_gen_prolog_loop_niters): ...here.
27505         (vect_update_init_of_dr): Take a code argument that specifies whether
27506         the adjustment should be added or subtracted.
27507         (vect_update_init_of_drs): Likewise.
27508         (vect_prepare_for_masked_peels): New function.
27509         (vect_do_peeling): Skip prologue peeling if we're using a mask
27510         instead.  Update call to vect_update_inits_of_drs.
27511         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
27512         mask_skip_niters.
27513         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
27514         alignment.  Do not include the number of peeled iterations in
27515         the minimum threshold in that case.
27516         (vectorizable_induction): Adjust the start value down by
27517         LOOP_VINFO_MASK_SKIP_NITERS iterations.
27518         (vect_transform_loop): Call vect_prepare_for_masked_peels.
27519         Take the number of skipped iterations into account when calculating
27520         the loop bounds.
27521         * tree-vect-stmts.c (vect_gen_while_not): New function.
27523 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27524             Alan Hayward  <alan.hayward@arm.com>
27525             David Sherwood  <david.sherwood@arm.com>
27527         * doc/sourcebuild.texi (vect_fully_masked): Document.
27528         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
27529         default value to 0.
27530         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
27531         split out from...
27532         (vect_analyze_loop_2): ...here. Don't check the vectorization
27533         factor against the number of loop iterations if the loop is
27534         fully-masked.
27536 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27537             Alan Hayward  <alan.hayward@arm.com>
27538             David Sherwood  <david.sherwood@arm.com>
27540         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
27541         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
27542         (dump_groups): Update accordingly.
27543         (iv_use::mem_type): New member variable.
27544         (address_p): New function.
27545         (record_use): Add a mem_type argument and initialize the new
27546         mem_type field.
27547         (record_group_use): Add a mem_type argument.  Use address_p.
27548         Remove obsolete null checks of base_object.  Update call to record_use.
27549         (find_interesting_uses_op): Update call to record_group_use.
27550         (find_interesting_uses_cond): Likewise.
27551         (find_interesting_uses_address): Likewise.
27552         (get_mem_type_for_internal_fn): New function.
27553         (find_address_like_use): Likewise.
27554         (find_interesting_uses_stmt): Try find_address_like_use before
27555         calling find_interesting_uses_op.
27556         (addr_offset_valid_p): Use the iv mem_type field as the type
27557         of the addressed memory.
27558         (add_autoinc_candidates): Likewise.
27559         (get_address_cost): Likewise.
27560         (split_small_address_groups_p): Use address_p.
27561         (split_address_groups): Likewise.
27562         (add_iv_candidate_for_use): Likewise.
27563         (autoinc_possible_for_pair): Likewise.
27564         (rewrite_groups): Likewise.
27565         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
27566         (determine_group_iv_cost): Update after split of USE_ADDRESS.
27567         (get_alias_ptr_type_for_ptr_address): New function.
27568         (rewrite_use_address): Rewrite address uses in calls that were
27569         identified by find_address_like_use.
27571 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27572             Alan Hayward  <alan.hayward@arm.com>
27573             David Sherwood  <david.sherwood@arm.com>
27575         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
27576         TARGET_MEM_REFs.
27577         * gimple-expr.h (is_gimple_addressable: Likewise.
27578         * gimple-expr.c (is_gimple_address): Likewise.
27579         * internal-fn.c (expand_call_mem_ref): New function.
27580         (expand_mask_load_optab_fn): Use it.
27581         (expand_mask_store_optab_fn): Likewise.
27583 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27584             Alan Hayward  <alan.hayward@arm.com>
27585             David Sherwood  <david.sherwood@arm.com>
27587         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
27588         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
27589         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
27590         (cond_umax@var{mode}): Document.
27591         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
27592         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
27593         (cond_umin_optab, cond_umax_optab): New optabs.
27594         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
27595         (COND_IOR, COND_XOR): New internal functions.
27596         * internal-fn.h (get_conditional_internal_fn): Declare.
27597         * internal-fn.c (cond_binary_direct): New macro.
27598         (expand_cond_binary_optab_fn): Likewise.
27599         (direct_cond_binary_optab_supported_p): Likewise.
27600         (get_conditional_internal_fn): New function.
27601         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
27602         Cope with reduction statements that are vectorized as calls rather
27603         than assignments.
27604         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
27605         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
27606         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
27607         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
27608         (UNSPEC_COND_EOR): New unspecs.
27609         (optab): Add mappings for them.
27610         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
27611         (sve_int_op, sve_fp_op): New int attributes.
27613 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27614             Alan Hayward  <alan.hayward@arm.com>
27615             David Sherwood  <david.sherwood@arm.com>
27617         * optabs.def (while_ult_optab): New optab.
27618         * doc/md.texi (while_ult@var{m}@var{n}): Document.
27619         * internal-fn.def (WHILE_ULT): New internal function.
27620         * internal-fn.h (direct_internal_fn_supported_p): New override
27621         that takes two types as argument.
27622         * internal-fn.c (while_direct): New macro.
27623         (expand_while_optab_fn): New function.
27624         (convert_optab_supported_p): Likewise.
27625         (direct_while_optab_supported_p): New macro.
27626         * wide-int.h (wi::udiv_ceil): New function.
27627         * tree-vectorizer.h (rgroup_masks): New structure.
27628         (vec_loop_masks): New typedef.
27629         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
27630         and fully_masked_p.
27631         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
27632         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
27633         (vect_max_vf): New function.
27634         (slpeel_make_loop_iterate_ntimes): Delete.
27635         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
27636         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
27637         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
27638         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
27639         internal-fn.h, stor-layout.h and optabs-query.h.
27640         (vect_set_loop_mask): New function.
27641         (add_preheader_seq): Likewise.
27642         (add_header_seq): Likewise.
27643         (interleave_supported_p): Likewise.
27644         (vect_maybe_permute_loop_masks): Likewise.
27645         (vect_set_loop_masks_directly): Likewise.
27646         (vect_set_loop_condition_masked): Likewise.
27647         (vect_set_loop_condition_unmasked): New function, split out from
27648         slpeel_make_loop_iterate_ntimes.
27649         (slpeel_make_loop_iterate_ntimes): Rename to..
27650         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
27651         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
27652         (vect_do_peeling): Update call accordingly.
27653         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
27654         loops.
27655         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
27656         mask_compare_type, can_fully_mask_p and fully_masked_p.
27657         (release_vec_loop_masks): New function.
27658         (_loop_vec_info): Use it to free the loop masks.
27659         (can_produce_all_loop_masks_p): New function.
27660         (vect_get_max_nscalars_per_iter): Likewise.
27661         (vect_verify_full_masking): Likewise.
27662         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
27663         retries, and free the mask rgroups before retrying.  Check loop-wide
27664         reasons for disallowing fully-masked loops.  Make the final decision
27665         about whether use a fully-masked loop or not.
27666         (vect_estimate_min_profitable_iters): Do not assume that peeling
27667         for the number of iterations will be needed for fully-masked loops.
27668         (vectorizable_reduction): Disable fully-masked loops.
27669         (vectorizable_live_operation): Likewise.
27670         (vect_halve_mask_nunits): New function.
27671         (vect_double_mask_nunits): Likewise.
27672         (vect_record_loop_mask): Likewise.
27673         (vect_get_loop_mask): Likewise.
27674         (vect_transform_loop): Handle the case in which the final loop
27675         iteration might handle a partial vector.  Call vect_set_loop_condition
27676         instead of slpeel_make_loop_iterate_ntimes.
27677         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
27678         (check_load_store_masking): New function.
27679         (prepare_load_store_mask): Likewise.
27680         (vectorizable_store): Handle fully-masked loops.
27681         (vectorizable_load): Likewise.
27682         (supportable_widening_operation): Use vect_halve_mask_nunits for
27683         booleans.
27684         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
27685         (vect_gen_while): New function.
27686         * config/aarch64/aarch64.md (umax<mode>3): New expander.
27687         (aarch64_uqdec<mode>): New insn.
27689 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27690             Alan Hayward  <alan.hayward@arm.com>
27691             David Sherwood  <david.sherwood@arm.com>
27693         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
27694         (reduc_xor_scal_optab): New optabs.
27695         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
27696         (reduc_xor_scal_@var{m}): Document.
27697         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
27698         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
27699         internal functions.
27700         * fold-const-call.c (fold_const_call): Handle them.
27701         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
27702         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
27703         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
27704         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
27705         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
27706         (UNSPEC_XORV): New unspecs.
27707         (optab): Add entries for them.
27708         (BITWISEV): New int iterator.
27709         (bit_reduc_op): New int attributes.
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         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
27716         * internal-fn.def (VEC_SHL_INSERT): New internal function.
27717         * optabs.def (vec_shl_insert_optab): New optab.
27718         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
27719         (duplicate_and_interleave): Likewise.
27720         * tree-vect-loop.c: Include internal-fn.h.
27721         (neutral_op_for_slp_reduction): New function, split out from
27722         get_initial_defs_for_reduction.
27723         (get_initial_def_for_reduction): Handle option 2 for variable-length
27724         vectors by loading the neutral value into a vector and then shifting
27725         the initial value into element 0.
27726         (get_initial_defs_for_reduction): Replace the code argument with
27727         the neutral value calculated by neutral_op_for_slp_reduction.
27728         Use gimple_build_vector for constant-length vectors.
27729         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
27730         but the first group_size elements have a neutral value.
27731         Use duplicate_and_interleave otherwise.
27732         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
27733         Update call to get_initial_defs_for_reduction.  Handle SLP
27734         reductions for variable-length vectors by creating one vector
27735         result for each scalar result, with the elements associated
27736         with other scalar results stubbed out with the neutral value.
27737         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
27738         Require IFN_VEC_SHL_INSERT for double reductions on
27739         variable-length vectors, or SLP reductions that have
27740         a neutral value.  Require can_duplicate_and_interleave_p
27741         support for variable-length unchained SLP reductions if there
27742         is no neutral value, such as for MIN/MAX reductions.  Also require
27743         the number of vector elements to be a multiple of the number of
27744         SLP statements when doing variable-length unchained SLP reductions.
27745         Update call to vect_create_epilog_for_reduction.
27746         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
27747         and remove initial values.
27748         (duplicate_and_interleave): Make public.
27749         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
27750         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
27752 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27753             Alan Hayward  <alan.hayward@arm.com>
27754             David Sherwood  <david.sherwood@arm.com>
27756         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
27757         (can_duplicate_and_interleave_p): New function.
27758         (vect_get_and_check_slp_defs): Take the vector of statements
27759         rather than just the current one.  Remove excess parentheses.
27760         Restriction rejectinon of vect_constant_def and vect_external_def
27761         for variable-length vectors to boolean types, or types for which
27762         can_duplicate_and_interleave_p is false.
27763         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
27764         (duplicate_and_interleave): New function.
27765         (vect_get_constant_vectors): Use gimple_build_vector for
27766         constant-length vectors and suitable variable-length constant
27767         vectors.  Use duplicate_and_interleave for other variable-length
27768         vectors.  Don't defer the update when inserting new statements.
27770 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27771             Alan Hayward  <alan.hayward@arm.com>
27772             David Sherwood  <david.sherwood@arm.com>
27774         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
27775         min_profitable_iters doesn't go negative.
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         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
27782         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
27783         * optabs.def (vec_mask_load_lanes_optab): New optab.
27784         (vec_mask_store_lanes_optab): Likewise.
27785         * internal-fn.def (MASK_LOAD_LANES): New internal function.
27786         (MASK_STORE_LANES): Likewise.
27787         * internal-fn.c (mask_load_lanes_direct): New macro.
27788         (mask_store_lanes_direct): Likewise.
27789         (expand_mask_load_optab_fn): Handle masked operations.
27790         (expand_mask_load_lanes_optab_fn): New macro.
27791         (expand_mask_store_optab_fn): Handle masked operations.
27792         (expand_mask_store_lanes_optab_fn): New macro.
27793         (direct_mask_load_lanes_optab_supported_p): Likewise.
27794         (direct_mask_store_lanes_optab_supported_p): Likewise.
27795         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
27796         parameter.
27797         (vect_load_lanes_supported): Likewise.
27798         * tree-vect-data-refs.c (strip_conversion): New function.
27799         (can_group_stmts_p): Likewise.
27800         (vect_analyze_data_ref_accesses): Use it instead of checking
27801         for a pair of assignments.
27802         (vect_store_lanes_supported): Take a masked_p parameter.
27803         (vect_load_lanes_supported): Likewise.
27804         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
27805         vect_store_lanes_supported and vect_load_lanes_supported.
27806         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
27807         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
27808         parameter.  Don't allow gaps for masked accesses.
27809         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
27810         and vect_load_lanes_supported.
27811         (get_load_store_type): Take a masked_p parameter and update
27812         call to get_group_load_store_type.
27813         (vectorizable_store): Update call to get_load_store_type.
27814         Handle IFN_MASK_STORE_LANES.
27815         (vectorizable_load): Update call to get_load_store_type.
27816         Handle IFN_MASK_LOAD_LANES.
27818 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27819             Alan Hayward  <alan.hayward@arm.com>
27820             David Sherwood  <david.sherwood@arm.com>
27822         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
27823         modes for SVE.
27824         * config/aarch64/aarch64-protos.h
27825         (aarch64_sve_struct_memory_operand_p): Declare.
27826         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
27827         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
27828         (VPRED, vpred): Handle SVE structure modes.
27829         * config/aarch64/constraints.md (Utx): New constraint.
27830         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
27831         (aarch64_sve_struct_nonimmediate_operand): New predicates.
27832         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
27833         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
27834         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
27835         structure modes.  Split into pieces after RA.
27836         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
27837         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
27838         New patterns.
27839         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
27840         SVE structure modes.
27841         (aarch64_classify_address): Likewise.
27842         (sizetochar): Move earlier in file.
27843         (aarch64_print_operand): Handle SVE register lists.
27844         (aarch64_array_mode): New function.
27845         (aarch64_sve_struct_memory_operand_p): Likewise.
27846         (TARGET_ARRAY_MODE): Redefine.
27848 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27849             Alan Hayward  <alan.hayward@arm.com>
27850             David Sherwood  <david.sherwood@arm.com>
27852         * target.def (array_mode): New target hook.
27853         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
27854         * doc/tm.texi: Regenerate.
27855         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
27856         * hooks.c (hook_optmode_mode_uhwi_none): New function.
27857         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
27858         targetm.array_mode.
27859         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
27860         type sizes.
27862 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27863             Alan Hayward  <alan.hayward@arm.com>
27864             David Sherwood  <david.sherwood@arm.com>
27866         * fold-const.c (fold_binary_loc): Check the argument types
27867         rather than the result type when testing for a vector operation.
27869 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27871         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
27872         * doc/tm.texi: Regenerate.
27874 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
27875             Alan Hayward  <alan.hayward@arm.com>
27876             David Sherwood  <david.sherwood@arm.com>
27878         * doc/invoke.texi (-msve-vector-bits=): Document new option.
27879         (sve): Document new AArch64 extension.
27880         * doc/md.texi (w): Extend the description of the AArch64
27881         constraint to include SVE vectors.
27882         (Upl, Upa): Document new AArch64 predicate constraints.
27883         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
27884         enum.
27885         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
27886         (msve-vector-bits=): New option.
27887         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
27888         SVE when these are disabled.
27889         (sve): New extension.
27890         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
27891         modes.  Adjust their number of units based on aarch64_sve_vg.
27892         (MAX_BITSIZE_MODE_ANY_MODE): Define.
27893         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
27894         aarch64_addr_query_type.
27895         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
27896         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
27897         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
27898         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
27899         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
27900         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
27901         (aarch64_simd_imm_zero_p): Delete.
27902         (aarch64_check_zero_based_sve_index_immediate): Declare.
27903         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
27904         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
27905         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
27906         (aarch64_sve_float_mul_immediate_p): Likewise.
27907         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
27908         rather than an rtx.
27909         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
27910         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
27911         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
27912         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
27913         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
27914         (aarch64_regmode_natural_size): Likewise.
27915         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
27916         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
27917         left one place.
27918         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
27919         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
27920         for VG and the SVE predicate registers.
27921         (V_ALIASES): Add a "z"-prefixed alias.
27922         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
27923         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
27924         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
27925         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
27926         (REG_CLASS_NAMES): Add entries for them.
27927         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
27928         and the predicate registers.
27929         (aarch64_sve_vg): Declare.
27930         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
27931         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
27932         (REGMODE_NATURAL_SIZE): Define.
27933         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
27934         SVE macros.
27935         * config/aarch64/aarch64.c: Include cfgrtl.h.
27936         (simd_immediate_info): Add a constructor for series vectors,
27937         and an associated step field.
27938         (aarch64_sve_vg): New variable.
27939         (aarch64_dbx_register_number): Handle VG and the predicate registers.
27940         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
27941         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
27942         (VEC_ANY_DATA, VEC_STRUCT): New constants.
27943         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
27944         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
27945         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
27946         (aarch64_get_mask_mode): New functions.
27947         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
27948         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
27949         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
27950         predicate modes and predicate registers.  Explicitly restrict
27951         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
27952         to store a vector mode if it is recognized by
27953         aarch64_classify_vector_mode.
27954         (aarch64_regmode_natural_size): New function.
27955         (aarch64_hard_regno_caller_save_mode): Return the original mode
27956         for predicates.
27957         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
27958         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
27959         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
27960         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
27961         functions.
27962         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
27963         does not overlap dest if the function is frame-related.  Handle
27964         SVE constants.
27965         (aarch64_split_add_offset): New function.
27966         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
27967         them aarch64_add_offset.
27968         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
27969         and update call to aarch64_sub_sp.
27970         (aarch64_add_cfa_expression): New function.
27971         (aarch64_expand_prologue): Pass extra temporary registers to the
27972         functions above.  Handle the case in which we need to emit new
27973         DW_CFA_expressions for registers that were originally saved
27974         relative to the stack pointer, but now have to be expressed
27975         relative to the frame pointer.
27976         (aarch64_output_mi_thunk): Pass extra temporary registers to the
27977         functions above.
27978         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
27979         IP0 and IP1 values for SVE frames.
27980         (aarch64_expand_vec_series): New function.
27981         (aarch64_expand_sve_widened_duplicate): Likewise.
27982         (aarch64_expand_sve_const_vector): Likewise.
27983         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
27984         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
27985         into the register, rather than emitting a SET directly.
27986         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
27987         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
27988         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
27989         (offset_9bit_signed_scaled_p): New functions.
27990         (aarch64_replicate_bitmask_imm): New function.
27991         (aarch64_bitmask_imm): Use it.
27992         (aarch64_cannot_force_const_mem): Reject expressions involving
27993         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
27994         (aarch64_classify_index): Handle SVE indices, by requiring
27995         a plain register index with a scale that matches the element size.
27996         (aarch64_classify_address): Handle SVE addresses.  Assert that
27997         the mode of the address is VOIDmode or an integer mode.
27998         Update call to aarch64_classify_symbol.
27999         (aarch64_classify_symbolic_expression): Update call to
28000         aarch64_classify_symbol.
28001         (aarch64_const_vec_all_in_range_p): New function.
28002         (aarch64_print_vector_float_operand): Likewise.
28003         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
28004         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
28005         and the FP immediates 1.0 and 0.5.
28006         (aarch64_print_address_internal): Handle SVE addresses.
28007         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
28008         (aarch64_regno_regclass): Handle predicate registers.
28009         (aarch64_secondary_reload): Handle big-endian reloads of SVE
28010         data modes.
28011         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
28012         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
28013         (aarch64_convert_sve_vector_bits): New function.
28014         (aarch64_override_options): Use it to handle -msve-vector-bits=.
28015         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
28016         rather than an rtx.
28017         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
28018         Handle SVE vector and predicate modes.  Accept VL-based constants
28019         that need only one temporary register, and VL offsets that require
28020         no temporary registers.
28021         (aarch64_conditional_register_usage): Mark the predicate registers
28022         as fixed if SVE isn't available.
28023         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
28024         Return true for SVE vector and predicate modes.
28025         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
28026         rather than an unsigned int.  Handle SVE modes.
28027         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
28028         SVE modes.
28029         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
28030         if SVE is enabled.
28031         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
28032         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
28033         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
28034         (aarch64_sve_float_mul_immediate_p): New functions.
28035         (aarch64_sve_valid_immediate): New function.
28036         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
28037         Explicitly reject structure modes.  Check for INDEX constants.
28038         Handle PTRUE and PFALSE constants.
28039         (aarch64_check_zero_based_sve_index_immediate): New function.
28040         (aarch64_simd_imm_zero_p): Delete.
28041         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
28042         vector modes.  Accept constants in the range of CNT[BHWD].
28043         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
28044         ask for an Advanced SIMD mode.
28045         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
28046         (aarch64_simd_vector_alignment): Handle SVE predicates.
28047         (aarch64_vectorize_preferred_vector_alignment): New function.
28048         (aarch64_simd_vector_alignment_reachable): Use it instead of
28049         the vector size.
28050         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
28051         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
28052         functions.
28053         (MAX_VECT_LEN): Delete.
28054         (expand_vec_perm_d): Add a vec_flags field.
28055         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
28056         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
28057         (aarch64_evpc_ext): Don't apply a big-endian lane correction
28058         for SVE modes.
28059         (aarch64_evpc_rev): Rename to...
28060         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
28061         (aarch64_evpc_rev_global): New function.
28062         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
28063         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
28064         MAX_VECT_LEN.
28065         (aarch64_evpc_sve_tbl): New function.
28066         (aarch64_expand_vec_perm_const_1): Update after rename of
28067         aarch64_evpc_rev.  Handle SVE permutes too, trying
28068         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
28069         than aarch64_evpc_tbl.
28070         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
28071         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
28072         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
28073         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
28074         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
28075         (aarch64_expand_sve_vcond): New functions.
28076         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
28077         of aarch64_vector_mode_p.
28078         (aarch64_dwarf_poly_indeterminate_value): New function.
28079         (aarch64_compute_pressure_classes): Likewise.
28080         (aarch64_can_change_mode_class): Likewise.
28081         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
28082         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
28083         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
28084         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
28085         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
28086         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
28087         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
28088         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
28089         constraints.
28090         (Dn, Dl, Dr): Accept const as well as const_vector.
28091         (Dz): Likewise.  Compare against CONST0_RTX.
28092         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
28093         of "vector" where appropriate.
28094         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
28095         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
28096         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
28097         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
28098         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
28099         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
28100         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
28101         (v_int_equiv): Extend to SVE modes.
28102         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
28103         mode attributes.
28104         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
28105         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
28106         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
28107         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
28108         (SVE_COND_FP_CMP): New int iterators.
28109         (perm_hilo): Handle the new unpack unspecs.
28110         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
28111         attributes.
28112         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
28113         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
28114         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
28115         (aarch64_equality_operator, aarch64_constant_vector_operand)
28116         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
28117         (aarch64_sve_nonimmediate_operand): Likewise.
28118         (aarch64_sve_general_operand): Likewise.
28119         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
28120         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
28121         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
28122         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
28123         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
28124         (aarch64_sve_float_arith_immediate): Likewise.
28125         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
28126         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
28127         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
28128         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
28129         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
28130         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
28131         (aarch64_sve_float_arith_operand): Likewise.
28132         (aarch64_sve_float_arith_with_sub_operand): Likewise.
28133         (aarch64_sve_float_mul_operand): Likewise.
28134         (aarch64_sve_vec_perm_operand): Likewise.
28135         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
28136         (aarch64_mov_operand): Accept const_poly_int and const_vector.
28137         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
28138         as well as const_vector.
28139         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
28140         in file.  Use CONST0_RTX and CONSTM1_RTX.
28141         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
28142         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
28143         Use aarch64_simd_imm_zero.
28144         * config/aarch64/aarch64-sve.md: New file.
28145         * config/aarch64/aarch64.md: Include it.
28146         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
28147         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
28148         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
28149         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
28150         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
28151         (sve): New attribute.
28152         (enabled): Disable instructions with the sve attribute unless
28153         TARGET_SVE.
28154         (movqi, movhi): Pass CONST_POLY_INT operaneds through
28155         aarch64_expand_mov_immediate.
28156         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
28157         CNT[BHSD] immediates.
28158         (movti): Split CONST_POLY_INT moves into two halves.
28159         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
28160         Split additions that need a temporary here if the destination
28161         is the stack pointer.
28162         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
28163         (*add<mode>3_poly_1): New instruction.
28164         (set_clobber_cc): New expander.
28166 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28168         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
28169         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
28170         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
28171         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
28172         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
28173         Change innermode from fixed_mode_size to machine_mode.
28174         (simplify_subreg): Update call accordingly.  Handle a constant-sized
28175         subreg of a variable-length CONST_VECTOR.
28177 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28178             Alan Hayward  <alan.hayward@arm.com>
28179             David Sherwood  <david.sherwood@arm.com>
28181         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
28182         (add_offset_to_base): New function, split out from...
28183         (create_mem_ref): ...here.  When handling a scale other than 1,
28184         check first whether the address is valid without the offset.
28185         Add it into the base if so, leaving the index and scale as-is.
28187 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
28189         PR c++/83778
28190         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
28191         fold_for_warn before checking if arg2 is INTEGER_CST.
28193 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
28195         * config/rs6000/predicates.md (load_multiple_operation): Delete.
28196         (store_multiple_operation): Delete.
28197         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
28198         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
28199         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
28200         guarded by TARGET_STRING.
28201         (rs6000_output_load_multiple): Delete.
28202         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
28203         OPTION_MASK_STRING / TARGET_STRING handling.
28204         (print_operand) <'N', 'O'>: Add comment that these are unused now.
28205         (const rs6000_opt_masks) <"string">: Change mask to 0.
28206         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
28207         (MASK_STRING): Delete.
28208         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
28209         parts.  Simplify.
28210         (load_multiple): Delete.
28211         (*ldmsi8): Delete.
28212         (*ldmsi7): Delete.
28213         (*ldmsi6): Delete.
28214         (*ldmsi5): Delete.
28215         (*ldmsi4): Delete.
28216         (*ldmsi3): Delete.
28217         (store_multiple): Delete.
28218         (*stmsi8): Delete.
28219         (*stmsi7): Delete.
28220         (*stmsi6): Delete.
28221         (*stmsi5): Delete.
28222         (*stmsi4): Delete.
28223         (*stmsi3): Delete.
28224         (movmemsi_8reg): Delete.
28225         (corresponding unnamed define_insn): Delete.
28226         (movmemsi_6reg): Delete.
28227         (corresponding unnamed define_insn): Delete.
28228         (movmemsi_4reg): Delete.
28229         (corresponding unnamed define_insn): Delete.
28230         (movmemsi_2reg): Delete.
28231         (corresponding unnamed define_insn): Delete.
28232         (movmemsi_1reg): Delete.
28233         (corresponding unnamed define_insn): Delete.
28234         * config/rs6000/rs6000.opt (mno-string): New.
28235         (mstring): Replace by deprecation warning stub.
28236         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
28238 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
28240         * regrename.c (regrename_do_replace): If replacing the same
28241         reg multiple times, try to reuse last created gen_raw_REG.
28243         PR debug/81155
28244         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
28245         main to workaround a bug in GDB.
28247 2018-01-12  Tom de Vries  <tom@codesourcery.com>
28249         PR target/83737
28250         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
28252 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
28254         PR rtl-optimization/80481
28255         * ira-color.c (get_cap_member): New function.
28256         (allocnos_conflict_by_live_ranges_p): Use it.
28257         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
28258         (setup_slot_coalesced_allocno_live_ranges): Ditto.
28260 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
28262         PR target/83628
28263         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
28264         (*saddl_se_1): Ditto.
28265         (*ssubsi_1): Ditto.
28266         (*ssubl_se_1): Ditto.
28268 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
28270         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
28271         rather than wi::to_widest for DR_INITs.
28272         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
28273         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
28274         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
28275         INTEGER_CSTs.
28276         (vect_analyze_group_access_1): Note that here.
28278 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
28280         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
28281         polynomial type sizes.
28283 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
28285         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
28286         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
28287         (gimple_add_tmp_var): Likewise.
28289 2018-01-12  Martin Liska  <mliska@suse.cz>
28291         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
28292         (gimple_alloc_sizes): Likewise.
28293         (dump_gimple_statistics): Use PRIu64 in printf format.
28294         * gimple.h: Change uint64_t to int.
28296 2018-01-12  Martin Liska  <mliska@suse.cz>
28298         * tree-core.h: Use uint64_t instead of int.
28299         * tree.c (tree_node_counts): Likewise.
28300         (tree_node_sizes): Likewise.
28301         (dump_tree_statistics): Use PRIu64 in printf format.
28303 2018-01-12  Martin Liska  <mliska@suse.cz>
28305         * Makefile.in: As qsort_chk is implemented in vec.c, add
28306         vec.o to linkage of gencfn-macros.
28307         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
28308         passing the info to record_node_allocation_statistics.
28309         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
28310         and pass the info.
28311         * ggc-common.c (struct ggc_usage): Add operator== and use
28312         it in operator< and compare function.
28313         * mem-stats.h (struct mem_usage): Likewise.
28314         * vec.c (struct vec_usage): Remove operator< and compare
28315         function. Can be simply inherited.
28317 2018-01-12  Martin Jambor  <mjambor@suse.cz>
28319         PR target/81616
28320         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
28321         * tree-ssa-math-opts.c: Include domwalk.h.
28322         (convert_mult_to_fma_1): New function.
28323         (fma_transformation_info): New type.
28324         (fma_deferring_state): Likewise.
28325         (cancel_fma_deferring): New function.
28326         (result_of_phi): Likewise.
28327         (last_fma_candidate_feeds_initial_phi): Likewise.
28328         (convert_mult_to_fma): Added deferring logic, split actual
28329         transformation to convert_mult_to_fma_1.
28330         (math_opts_dom_walker): New type.
28331         (math_opts_dom_walker::after_dom_children): New method, body moved
28332         here from pass_optimize_widening_mul::execute, added deferring logic
28333         bits.
28334         (pass_optimize_widening_mul::execute): Moved most of code to
28335         math_opts_dom_walker::after_dom_children.
28336         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
28337         * config/i386/i386.c (ix86_option_override_internal): Added
28338         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
28340 2018-01-12  Richard Biener  <rguenther@suse.de>
28342         PR debug/83157
28343         * dwarf2out.c (gen_variable_die): Do not reset old_die for
28344         inline instance vars.
28346 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
28348         PR target/81819
28349         * config/rx/rx.c (rx_is_restricted_memory_address):
28350         Handle SUBREG case.
28352 2018-01-12  Richard Biener  <rguenther@suse.de>
28354         PR tree-optimization/80846
28355         * target.def (split_reduction): New target hook.
28356         * targhooks.c (default_split_reduction): New function.
28357         * targhooks.h (default_split_reduction): Declare.
28358         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
28359         target requests first reduce vectors by combining low and high
28360         parts.
28361         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
28362         (get_vectype_for_scalar_type_and_size): Export.
28363         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
28364         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
28365         * doc/tm.texi: Regenerate.
28366         * config/i386/i386.c (ix86_split_reduction): Implement
28367         TARGET_VECTORIZE_SPLIT_REDUCTION.
28369 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
28371         PR target/83368
28372         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
28373         in PIC mode except for TARGET_VXWORKS_RTP.
28374         * config/sparc/sparc.c: Include cfgrtl.h.
28375         (TARGET_INIT_PIC_REG): Define.
28376         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
28377         (sparc_pic_register_p): New predicate.
28378         (sparc_legitimate_address_p): Use it.
28379         (sparc_legitimize_pic_address): Likewise.
28380         (sparc_delegitimize_address): Likewise.
28381         (sparc_mode_dependent_address_p): Likewise.
28382         (gen_load_pcrel_sym): Remove 4th parameter.
28383         (load_got_register): Adjust call to above.  Remove obsolete stuff.
28384         (sparc_expand_prologue): Do not call load_got_register here.
28385         (sparc_flat_expand_prologue): Likewise.
28386         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
28387         (sparc_use_pseudo_pic_reg): New function.
28388         (sparc_init_pic_reg): Likewise.
28389         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
28390         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
28392 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
28394         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28395         Add item for branch_cost.
28397 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
28399         PR rtl-optimization/83565
28400         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
28401         not extend the result to a larger mode for rotate operations.
28402         (num_sign_bit_copies1): Likewise.
28404 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
28406         PR target/40411
28407         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
28408         -symbolic.
28409         Use values-Xc.o for -pedantic.
28410         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
28412 2018-01-12  Martin Liska  <mliska@suse.cz>
28414         PR ipa/83054
28415         * ipa-devirt.c (final_warning_record::grow_type_warnings):
28416         New function.
28417         (possible_polymorphic_call_targets): Use it.
28418         (ipa_devirt): Likewise.
28420 2018-01-12  Martin Liska  <mliska@suse.cz>
28422         * profile-count.h (enum profile_quality): Use 0 as invalid
28423         enum value of profile_quality.
28425 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
28427         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
28428         -mext-string options.
28430 2018-01-12  Richard Biener  <rguenther@suse.de>
28432         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
28433         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
28434         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
28435         Likewise.
28436         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
28438 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
28440         * configure.ac (--with-long-double-format): Add support for the
28441         configuration option to change the default long double format on
28442         PowerPC systems.
28443         * config.gcc (powerpc*-linux*-*): Likewise.
28444         * configure: Regenerate.
28445         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
28446         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
28447         used without modification.
28449 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28451         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
28452         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
28453         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
28454         MISC_BUILTIN_SPEC_BARRIER.
28455         (rs6000_init_builtins): Likewise.
28456         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
28457         enum value.
28458         (speculation_barrier): New define_insn.
28459         * doc/extend.texi: Document __builtin_speculation_barrier.
28461 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
28463         PR target/83203
28464         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
28465         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
28466         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
28467         iterators.
28468         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
28469         integral modes instead of "ss" and "sd".
28470         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
28471         vectors with 32-bit and 64-bit elements.
28472         (vecdupssescalarmodesuffix): New mode attribute.
28473         (vec_dup<mode>): Use it.
28475 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
28477         PR target/83330
28478         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
28479         frame if argument is passed on stack.
28481 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
28483         PR target/82682
28484         * ree.c (combine_reaching_defs): Optimize also
28485         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
28486         reg2=any_extend(exp); reg1=reg2;, formatting fix.
28488 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
28490         PR middle-end/83189
28491         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
28493 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
28495         PR middle-end/83718
28496         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
28497         after they are computed.
28499 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
28501         PR tree-optimization/83695
28502         * gimple-loop-linterchange.cc
28503         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
28504         reset cached scev information after interchange.
28505         (pass_linterchange::execute): Remove call to scev_reset_htab.
28507 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28509         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
28510         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
28511         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
28512         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
28513         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
28514         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
28515         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
28516         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
28517         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
28518         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
28519         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
28520         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
28521         (V_lane_reg): Likewise.
28522         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
28523         New define_expand.
28524         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
28525         (vfmal_lane_low<mode>_intrinsic,
28526         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
28527         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
28528         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
28529         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
28530         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
28531         vfmsl_lane_high<mode>_intrinsic): New define_insns.
28533 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28535         * config/arm/arm-cpus.in (fp16fml): New feature.
28536         (ALL_SIMD): Add fp16fml.
28537         (armv8.2-a): Add fp16fml as an option.
28538         (armv8.3-a): Likewise.
28539         (armv8.4-a): Add fp16fml as part of fp16.
28540         * config/arm/arm.h (TARGET_FP16FML): Define.
28541         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
28542         when appropriate.
28543         * config/arm/arm-modes.def (V2HF): Define.
28544         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
28545         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
28546         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
28547         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
28548         vfmsl_low, vfmsl_high): New set of builtins.
28549         * config/arm/iterators.md (PLUSMINUS): New code iterator.
28550         (vfml_op): New code attribute.
28551         (VFMLHALVES): New int iterator.
28552         (VFML, VFMLSEL): New mode attributes.
28553         (V_reg): Define mapping for V2HF.
28554         (V_hi, V_lo): New mode attributes.
28555         (VF_constraint): Likewise.
28556         (vfml_half, vfml_half_selector): New int attributes.
28557         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
28558         define_expand.
28559         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
28560         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
28561         New define_insn.
28562         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
28563         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
28564         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
28565         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
28566         documentation.
28567         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
28568         Document new effective target and option set.
28570 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28572         * config/arm/arm-cpus.in (armv8_4): New feature.
28573         (ARMv8_4a): New fgroup.
28574         (armv8.4-a): New arch.
28575         * config/arm/arm-tables.opt: Regenerate.
28576         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
28577         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
28578         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
28579         Add matching rules for -march=armv8.4-a and extensions.
28580         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
28582 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
28584         PR target/81821
28585         * config/rx/rx.md (BW): New mode attribute.
28586         (sync_lock_test_and_setsi): Add mode suffix to insn output.
28588 2018-01-11  Richard Biener  <rguenther@suse.de>
28590         PR tree-optimization/83435
28591         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
28592         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
28593         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
28595 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
28596             Alan Hayward  <alan.hayward@arm.com>
28597             David Sherwood  <david.sherwood@arm.com>
28599         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
28600         field.
28601         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
28602         (aarch64_print_address_internal): Use it to check for a zero offset.
28604 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
28605             Alan Hayward  <alan.hayward@arm.com>
28606             David Sherwood  <david.sherwood@arm.com>
28608         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
28609         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
28610         Return a poly_int64 rather than a HOST_WIDE_INT.
28611         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
28612         rather than a HOST_WIDE_INT.
28613         * config/aarch64/aarch64.h (aarch64_frame): Protect with
28614         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
28615         hard_fp_offset, frame_size, initial_adjust, callee_offset and
28616         final_offset from HOST_WIDE_INT to poly_int64.
28617         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
28618         to_constant when getting the number of units in an Advanced SIMD
28619         mode.
28620         (aarch64_builtin_vectorized_function): Check for a constant number
28621         of units.
28622         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
28623         GET_MODE_SIZE.
28624         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
28625         attribute instead of GET_MODE_NUNITS.
28626         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
28627         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
28628         GET_MODE_SIZE for fixed-size registers.
28629         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
28630         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
28631         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
28632         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
28633         (aarch64_print_operand, aarch64_print_address_internal)
28634         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
28635         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
28636         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
28637         Handle polynomial GET_MODE_SIZE.
28638         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
28639         wider than SImode without modification.
28640         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
28641         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
28642         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
28643         passing and returning SVE modes.
28644         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
28645         rather than GEN_INT.
28646         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
28647         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
28648         (aarch64_allocate_and_probe_stack_space): Likewise.
28649         (aarch64_layout_frame): Cope with polynomial offsets.
28650         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
28651         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
28652         polynomial offsets.
28653         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
28654         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
28655         poly_int64 rather than a HOST_WIDE_INT.
28656         (aarch64_get_separate_components, aarch64_process_components)
28657         (aarch64_expand_prologue, aarch64_expand_epilogue)
28658         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
28659         (aarch64_anchor_offset): New function, split out from...
28660         (aarch64_legitimize_address): ...here.
28661         (aarch64_builtin_vectorization_cost): Handle polynomial
28662         TYPE_VECTOR_SUBPARTS.
28663         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
28664         GET_MODE_NUNITS.
28665         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
28666         number of elements from the PARALLEL rather than the mode.
28667         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
28668         rather than GET_MODE_BITSIZE.
28669         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
28670         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
28671         (aarch64_expand_vec_perm_const_1): Handle polynomial
28672         d->perm.length () and d->perm elements.
28673         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
28674         Apply to_constant to d->perm elements.
28675         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
28676         polynomial CONST_VECTOR_NUNITS.
28677         (aarch64_move_pointer): Take amount as a poly_int64 rather
28678         than an int.
28679         (aarch64_progress_pointer): Avoid temporary variable.
28680         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
28681         the mode attribute instead of GET_MODE.
28683 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
28684             Alan Hayward  <alan.hayward@arm.com>
28685             David Sherwood  <david.sherwood@arm.com>
28687         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
28688         x exists before using it.
28689         (aarch64_add_constant_internal): Rename to...
28690         (aarch64_add_offset_1): ...this.  Replace regnum with separate
28691         src and dest rtxes.  Handle the case in which they're different,
28692         including when the offset is zero.  Replace scratchreg with an rtx.
28693         Use 2 additions if there is no spare register into which we can
28694         move a 16-bit constant.
28695         (aarch64_add_constant): Delete.
28696         (aarch64_add_offset): Replace reg with separate src and dest
28697         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
28698         Use aarch64_add_offset_1.
28699         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
28700         an rtx rather than an int.  Take the delta as a poly_int64
28701         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
28702         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
28703         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
28704         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
28705         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
28706         and aarch64_add_sp.
28707         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
28708         aarch64_add_constant.
28710 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
28712         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
28713         Use scalar_float_mode.
28715 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
28717         * config/aarch64/aarch64-simd.md
28718         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
28719         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
28720         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
28721         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
28722         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
28723         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
28724         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
28725         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
28726         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
28727         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
28729 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
28731         PR target/83514
28732         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
28733         targ_options->x_arm_arch_string is non NULL.
28735 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
28737         * config/aarch64/aarch64.h
28738         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
28740 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
28742         PR target/82096
28743         * expmed.c (emit_store_flag_force): Swap if const op0
28744         and change VOIDmode to mode of op0.
28746 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
28748         PR rtl-optimization/83761
28749         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
28750         than bytes to mode_for_size.
28752 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
28754         PR middle-end/83189
28755         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
28756         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
28757         profile.
28759 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
28761         PR middle-end/83575
28762         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
28763         when in layout mode.
28764         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
28765         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
28766         partition fixup.
28768 2018-01-10  Michael Collison  <michael.collison@arm.com>
28770         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
28771         * config/aarch64/aarch64-option-extension.def: Add
28772         AARCH64_OPT_EXTENSION of 'fp16fml'.
28773         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
28774         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
28775         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
28776         * config/aarch64/constraints.md (Ui7): New constraint.
28777         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
28778         (VFMLA_SEL_W): Ditto.
28779         (f16quad): Ditto.
28780         (f16mac1): Ditto.
28781         (VFMLA16_LOW): New int iterator.
28782         (VFMLA16_HIGH): Ditto.
28783         (UNSPEC_FMLAL): New unspec.
28784         (UNSPEC_FMLSL): Ditto.
28785         (UNSPEC_FMLAL2): Ditto.
28786         (UNSPEC_FMLSL2): Ditto.
28787         (f16mac): New code attribute.
28788         * config/aarch64/aarch64-simd-builtins.def
28789         (aarch64_fmlal_lowv2sf): Ditto.
28790         (aarch64_fmlsl_lowv2sf): Ditto.
28791         (aarch64_fmlalq_lowv4sf): Ditto.
28792         (aarch64_fmlslq_lowv4sf): Ditto.
28793         (aarch64_fmlal_highv2sf): Ditto.
28794         (aarch64_fmlsl_highv2sf): Ditto.
28795         (aarch64_fmlalq_highv4sf): Ditto.
28796         (aarch64_fmlslq_highv4sf): Ditto.
28797         (aarch64_fmlal_lane_lowv2sf): Ditto.
28798         (aarch64_fmlsl_lane_lowv2sf): Ditto.
28799         (aarch64_fmlal_laneq_lowv2sf): Ditto.
28800         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
28801         (aarch64_fmlalq_lane_lowv4sf): Ditto.
28802         (aarch64_fmlsl_lane_lowv4sf): Ditto.
28803         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
28804         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
28805         (aarch64_fmlal_lane_highv2sf): Ditto.
28806         (aarch64_fmlsl_lane_highv2sf): Ditto.
28807         (aarch64_fmlal_laneq_highv2sf): Ditto.
28808         (aarch64_fmlsl_laneq_highv2sf): Ditto.
28809         (aarch64_fmlalq_lane_highv4sf): Ditto.
28810         (aarch64_fmlsl_lane_highv4sf): Ditto.
28811         (aarch64_fmlalq_laneq_highv4sf): Ditto.
28812         (aarch64_fmlsl_laneq_highv4sf): Ditto.
28813         * config/aarch64/aarch64-simd.md:
28814         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
28815         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
28816         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
28817         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
28818         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
28819         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
28820         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
28821         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
28822         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
28823         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
28824         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
28825         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
28826         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
28827         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
28828         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
28829         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
28830         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
28831         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
28832         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
28833         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
28834         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
28835         (vfmlsl_low_u32): Ditto.
28836         (vfmlalq_low_u32): Ditto.
28837         (vfmlslq_low_u32): Ditto.
28838         (vfmlal_high_u32): Ditto.
28839         (vfmlsl_high_u32): Ditto.
28840         (vfmlalq_high_u32): Ditto.
28841         (vfmlslq_high_u32): Ditto.
28842         (vfmlal_lane_low_u32): Ditto.
28843         (vfmlsl_lane_low_u32): Ditto.
28844         (vfmlal_laneq_low_u32): Ditto.
28845         (vfmlsl_laneq_low_u32): Ditto.
28846         (vfmlalq_lane_low_u32): Ditto.
28847         (vfmlslq_lane_low_u32): Ditto.
28848         (vfmlalq_laneq_low_u32): Ditto.
28849         (vfmlslq_laneq_low_u32): Ditto.
28850         (vfmlal_lane_high_u32): Ditto.
28851         (vfmlsl_lane_high_u32): Ditto.
28852         (vfmlal_laneq_high_u32): Ditto.
28853         (vfmlsl_laneq_high_u32): Ditto.
28854         (vfmlalq_lane_high_u32): Ditto.
28855         (vfmlslq_lane_high_u32): Ditto.
28856         (vfmlalq_laneq_high_u32): Ditto.
28857         (vfmlslq_laneq_high_u32): Ditto.
28858         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
28859         (AARCH64_FL_FOR_ARCH8_4): New.
28860         (AARCH64_ISA_F16FML): New ISA flag.
28861         (TARGET_F16FML): New feature flag for fp16fml.
28862         (doc/invoke.texi): Document new fp16fml option.
28864 2018-01-10  Michael Collison  <michael.collison@arm.com>
28866         * config/aarch64/aarch64-builtins.c:
28867         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
28868         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
28869         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
28870         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
28871         (AARCH64_ISA_SHA3): New ISA flag.
28872         (TARGET_SHA3): New feature flag for sha3.
28873         * config/aarch64/iterators.md (sha512_op): New int attribute.
28874         (CRYPTO_SHA512): New int iterator.
28875         (UNSPEC_SHA512H): New unspec.
28876         (UNSPEC_SHA512H2): Ditto.
28877         (UNSPEC_SHA512SU0): Ditto.
28878         (UNSPEC_SHA512SU1): Ditto.
28879         * config/aarch64/aarch64-simd-builtins.def
28880         (aarch64_crypto_sha512hqv2di): New builtin.
28881         (aarch64_crypto_sha512h2qv2di): Ditto.
28882         (aarch64_crypto_sha512su0qv2di): Ditto.
28883         (aarch64_crypto_sha512su1qv2di): Ditto.
28884         (aarch64_eor3qv8hi): Ditto.
28885         (aarch64_rax1qv2di): Ditto.
28886         (aarch64_xarqv2di): Ditto.
28887         (aarch64_bcaxqv8hi): Ditto.
28888         * config/aarch64/aarch64-simd.md:
28889         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
28890         (aarch64_crypto_sha512su0qv2di): Ditto.
28891         (aarch64_crypto_sha512su1qv2di): Ditto.
28892         (aarch64_eor3qv8hi): Ditto.
28893         (aarch64_rax1qv2di): Ditto.
28894         (aarch64_xarqv2di): Ditto.
28895         (aarch64_bcaxqv8hi): Ditto.
28896         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
28897         (vsha512h2q_u64): Ditto.
28898         (vsha512su0q_u64): Ditto.
28899         (vsha512su1q_u64): Ditto.
28900         (veor3q_u16): Ditto.
28901         (vrax1q_u64): Ditto.
28902         (vxarq_u64): Ditto.
28903         (vbcaxq_u16): Ditto.
28904         * config/arm/types.md (crypto_sha512): New type attribute.
28905         (crypto_sha3): Ditto.
28906         (doc/invoke.texi): Document new sha3 option.
28908 2018-01-10  Michael Collison  <michael.collison@arm.com>
28910         * config/aarch64/aarch64-builtins.c:
28911         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
28912         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
28913         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
28914         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
28915         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
28916         (AARCH64_ISA_SM4): New ISA flag.
28917         (TARGET_SM4): New feature flag for sm4.
28918         * config/aarch64/aarch64-simd-builtins.def
28919         (aarch64_sm3ss1qv4si): Ditto.
28920         (aarch64_sm3tt1aq4si): Ditto.
28921         (aarch64_sm3tt1bq4si): Ditto.
28922         (aarch64_sm3tt2aq4si): Ditto.
28923         (aarch64_sm3tt2bq4si): Ditto.
28924         (aarch64_sm3partw1qv4si): Ditto.
28925         (aarch64_sm3partw2qv4si): Ditto.
28926         (aarch64_sm4eqv4si): Ditto.
28927         (aarch64_sm4ekeyqv4si): Ditto.
28928         * config/aarch64/aarch64-simd.md:
28929         (aarch64_sm3ss1qv4si): Ditto.
28930         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
28931         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
28932         (aarch64_sm4eqv4si): Ditto.
28933         (aarch64_sm4ekeyqv4si): Ditto.
28934         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
28935         (sm3part_op): Ditto.
28936         (CRYPTO_SM3TT): Ditto.
28937         (CRYPTO_SM3PART): Ditto.
28938         (UNSPEC_SM3SS1): New unspec.
28939         (UNSPEC_SM3TT1A): Ditto.
28940         (UNSPEC_SM3TT1B): Ditto.
28941         (UNSPEC_SM3TT2A): Ditto.
28942         (UNSPEC_SM3TT2B): Ditto.
28943         (UNSPEC_SM3PARTW1): Ditto.
28944         (UNSPEC_SM3PARTW2): Ditto.
28945         (UNSPEC_SM4E): Ditto.
28946         (UNSPEC_SM4EKEY): Ditto.
28947         * config/aarch64/constraints.md (Ui2): New constraint.
28948         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
28949         * config/arm/types.md (crypto_sm3): New type attribute.
28950         (crypto_sm4): Ditto.
28951         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
28952         (vsm3tt1aq_u32): Ditto.
28953         (vsm3tt1bq_u32): Ditto.
28954         (vsm3tt2aq_u32): Ditto.
28955         (vsm3tt2bq_u32): Ditto.
28956         (vsm3partw1q_u32): Ditto.
28957         (vsm3partw2q_u32): Ditto.
28958         (vsm4eq_u32): Ditto.
28959         (vsm4ekeyq_u32): Ditto.
28960         (doc/invoke.texi): Document new sm4 option.
28962 2018-01-10  Michael Collison  <michael.collison@arm.com>
28964         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
28965         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
28966         (AARCH64_FL_FOR_ARCH8_4): New.
28967         (AARCH64_FL_V8_4): New flag.
28968         (doc/invoke.texi): Document new armv8.4-a option.
28970 2018-01-10  Michael Collison  <michael.collison@arm.com>
28972         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
28973         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
28974         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
28975         * config/aarch64/aarch64-option-extension.def: Add
28976         AARCH64_OPT_EXTENSION of 'sha2'.
28977         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
28978         (crypto): Disable sha2 and aes if crypto disabled.
28979         (crypto): Enable aes and sha2 if enabled.
28980         (simd): Disable sha2 and aes if simd disabled.
28981         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
28982         New flags.
28983         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
28984         (TARGET_SHA2): New feature flag for sha2.
28985         (TARGET_AES): New feature flag for aes.
28986         * config/aarch64/aarch64-simd.md:
28987         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
28988         conditional on TARGET_AES.
28989         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
28990         (aarch64_crypto_sha1hsi): Make pattern conditional
28991         on TARGET_SHA2.
28992         (aarch64_crypto_sha1hv4si): Ditto.
28993         (aarch64_be_crypto_sha1hv4si): Ditto.
28994         (aarch64_crypto_sha1su1v4si): Ditto.
28995         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
28996         (aarch64_crypto_sha1su0v4si): Ditto.
28997         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
28998         (aarch64_crypto_sha256su0v4si): Ditto.
28999         (aarch64_crypto_sha256su1v4si): Ditto.
29000         (doc/invoke.texi): Document new aes and sha2 options.
29002 2018-01-10  Martin Sebor  <msebor@redhat.com>
29004         PR tree-optimization/83781
29005         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
29006         as string arrays.
29008 2018-01-11  Martin Sebor  <msebor@gmail.com>
29009             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29011         PR tree-optimization/83501
29012         PR tree-optimization/81703
29014         * tree-ssa-strlen.c (get_string_cst): Rename...
29015         (get_string_len): ...to this.  Handle global constants.
29016         (handle_char_store): Adjust.
29018 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
29019             Jim Wilson  <jimw@sifive.com>
29021         * config/riscv/riscv-protos.h (riscv_output_return): New.
29022         * config/riscv/riscv.c (struct machine_function): New naked_p field.
29023         (riscv_attribute_table, riscv_output_return),
29024         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
29025         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
29026         (riscv_compute_frame_info): Only compute frame->mask if not a naked
29027         function.
29028         (riscv_expand_prologue): Add early return for naked function.
29029         (riscv_expand_epilogue): Likewise.
29030         (riscv_function_ok_for_sibcall): Return false for naked function.
29031         (riscv_set_current_function): New.
29032         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
29033         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
29034         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
29035         * doc/extend.texi (RISC-V Function Attributes): New.
29037 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
29039         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
29040         check for 128-bit long double before checking TCmode.
29041         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
29042         128-bit long doubles before checking TFmode or TCmode.
29043         (FLOAT128_IBM_P): Likewise.
29045 2018-01-10  Martin Sebor  <msebor@redhat.com>
29047         PR tree-optimization/83671
29048         * builtins.c (c_strlen): Unconditionally return zero for the empty
29049         string.
29050         Use -Warray-bounds for warnings.
29051         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
29052         for non-constant array indices with COMPONENT_REF, arrays of
29053         arrays, and pointers to arrays.
29054         (gimple_fold_builtin_strlen): Determine and set length range for
29055         non-constant character arrays.
29057 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
29059         PR middle-end/81897
29060         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
29061         empty blocks.
29063 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
29065         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
29067 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
29069         PR target/83399
29070         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
29071         VECTOR_MEM_ALTIVEC_OR_VSX_P.
29072         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
29073         indexed_or_indirect_operand predicate.
29074         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
29075         (*vsx_le_perm_load_v8hi): Likewise.
29076         (*vsx_le_perm_load_v16qi): Likewise.
29077         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
29078         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
29079         (*vsx_le_perm_store_v8hi): Likewise.
29080         (*vsx_le_perm_store_v16qi): Likewise.
29081         (eight unnamed splitters): Likewise.
29083 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
29085         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
29086         * config/rs6000/emmintrin.h: Likewise.
29087         * config/rs6000/mmintrin.h: Likewise.
29088         * config/rs6000/xmmintrin.h: Likewise.
29090 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
29092         PR c++/43486
29093         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
29094         "public_flag".
29095         * tree.c (tree_nop_conversion): Return true for location wrapper
29096         nodes.
29097         (maybe_wrap_with_location): New function.
29098         (selftest::check_strip_nops): New function.
29099         (selftest::test_location_wrappers): New function.
29100         (selftest::tree_c_tests): Call it.
29101         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
29102         (maybe_wrap_with_location): New decl.
29103         (EXPR_LOCATION_WRAPPER_P): New macro.
29104         (location_wrapper_p): New inline function.
29105         (tree_strip_any_location_wrapper): New inline function.
29107 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
29109         PR target/83735
29110         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
29111         stack_realign_offset for the largest alignment of stack slot
29112         actually used.
29113         (ix86_find_max_used_stack_alignment): New function.
29114         (ix86_finalize_stack_frame_flags): Use it.  Set
29115         max_used_stack_alignment if we don't realign stack.
29116         * config/i386/i386.h (machine_function): Add
29117         max_used_stack_alignment.
29119 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
29121         * config/arm/arm.opt (-mbranch-cost): New option.
29122         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
29123         account.
29125 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
29127         PR target/83629
29128         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
29129         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
29131 2018-01-10  Richard Biener  <rguenther@suse.de>
29133         PR debug/83765
29134         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
29135         early out so it also covers the case where we have a non-NULL
29136         origin.
29138 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
29140         PR tree-optimization/83753
29141         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
29142         for non-strided grouped accesses if the number of elements is 1.
29144 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29146         PR target/81616
29147         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
29148         * i386.h (TARGET_USE_GATHER): Define.
29149         * x86-tune.def (X86_TUNE_USE_GATHER): New.
29151 2018-01-10  Martin Liska  <mliska@suse.cz>
29153         PR bootstrap/82831
29154         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
29155         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
29156         partitioning.
29157         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
29158         CLEANUP_NO_PARTITIONING is not set.
29160 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
29162         * doc/rtl.texi: Remove documentation of (const ...) wrappers
29163         for vectors, as a partial revert of r254296.
29164         * rtl.h (const_vec_p): Delete.
29165         (const_vec_duplicate_p): Don't test for vector CONSTs.
29166         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
29167         * expmed.c (make_tree): Likewise.
29169         Revert:
29170         * common.md (E, F): Use CONSTANT_P instead of checking for
29171         CONST_VECTOR.
29172         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
29173         checking for CONST_VECTOR.
29175 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
29177         PR middle-end/83575
29178         * predict.c (force_edge_cold): Handle in more sane way edges
29179         with no prediction.
29181 2018-01-09  Carl Love  <cel@us.ibm.com>
29183         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
29184         V4SI, V4SF types.
29185         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
29186         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
29187         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
29188         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
29189         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
29190         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
29191         * config/rs6000/rs6000-protos.h: Add extern defition for
29192         rs6000_generate_float2_double_code.
29193         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
29194         function.
29195         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
29196         (float2_v2df): Add define_expand.
29198 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
29200         PR target/83628
29201         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
29202         op_mode in the force_to_mode call.
29204 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
29206         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
29207         instead of checking each element individually.
29208         (aarch64_evpc_uzp): Likewise.
29209         (aarch64_evpc_zip): Likewise.
29210         (aarch64_evpc_ext): Likewise.
29211         (aarch64_evpc_rev): Likewise.
29212         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
29213         instead of checking each element individually.  Return true without
29214         generating rtl if
29215         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
29216         whether all selected elements come from the same input, instead of
29217         checking each element individually.  Remove calls to gen_rtx_REG,
29218         start_sequence and end_sequence and instead assert that no rtl is
29219         generated.
29221 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
29223         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
29224         order of HIGH and CONST checks.
29226 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
29228         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
29229         if the destination isn't an SSA_NAME.
29231 2018-01-09  Richard Biener  <rguenther@suse.de>
29233         PR tree-optimization/83668
29234         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
29235         move prologue...
29236         (canonicalize_loop_form): ... here, renamed from ...
29237         (canonicalize_loop_closed_ssa_form): ... this and amended to
29238         swap successor edges for loop exit blocks to make us use
29239         the RPO order we need for initial schedule generation.
29241 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
29243         PR tree-optimization/64811
29244         * match.pd: When optimizing comparisons with Inf, avoid
29245         introducing or losing exceptions from comparisons with NaN.
29247 2018-01-09  Martin Liska  <mliska@suse.cz>
29249         PR sanitizer/82517
29250         * asan.c (shadow_mem_size): Add gcc_assert.
29252 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
29254         Don't save registers in main().
29256         PR target/83738
29257         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
29258         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
29259         * config/avr/avr.c (avr_set_current_function): Don't error if
29260         naked, OS_task or OS_main are specified at the same time.
29261         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
29262         OS_main.
29263         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
29264         attribute.
29265         * common/config/avr/avr-common.c (avr_option_optimization_table):
29266         Switch on -mmain-is-OS_task for optimizing compilations.
29268 2018-01-09  Richard Biener  <rguenther@suse.de>
29270         PR tree-optimization/83572
29271         * graphite.c: Include cfganal.h.
29272         (graphite_transform_loops): Connect infinite loops to exit
29273         and remove fake edges at the end.
29275 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
29277         * ipa-inline.c (edge_badness): Revert accidental checkin.
29279 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
29281         PR ipa/80763
29282         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
29283         symbols; not inline clones.
29285 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
29287         PR target/83507
29288         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
29289         hard registers.  Formatting fixes.
29291         PR preprocessor/83722
29292         * gcc.c (try_generate_repro): Pass
29293         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
29294         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
29295         do_report_bug.
29297 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
29298             Kito Cheng  <kito.cheng@gmail.com>
29300         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
29301         (riscv_leaf_function_p): Delete.
29302         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
29304 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29306         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
29307         function.
29308         (do_ifelse): New function.
29309         (do_isel): New function.
29310         (do_sub3): New function.
29311         (do_add3): New function.
29312         (do_load_mask_compare): New function.
29313         (do_overlap_load_compare): New function.
29314         (expand_compare_loop): New function.
29315         (expand_block_compare): Call expand_compare_loop() when appropriate.
29316         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
29317         option description.
29318         (-mblock-compare-inline-loop-limit): New option.
29320 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29322         PR target/83677
29323         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
29324         Reverse order of second and third operands in first alternative.
29325         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
29326         of first and second elements in UNSPEC_VPERMR vector.
29327         (altivec_expand_vec_perm_le): Likewise.
29329 2018-01-08  Jeff Law  <law@redhat.com>
29331         PR rtl-optimizatin/81308
29332         * tree-switch-conversion.c (cfg_altered): New file scoped static.
29333         (process_switch): If group_case_labels makes a change, then set
29334         cfg_altered.
29335         (pass_convert_switch::execute): If a switch is converted, then
29336         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
29338         PR rtl-optimization/81308
29339         * recog.c (split_all_insns): Conditionally cleanup the CFG after
29340         splitting insns.
29342 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
29344         PR target/83663 - Revert r255946
29345         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
29346         generation for cases where splatting a value is not useful.
29347         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
29348         across a vec_duplicate and a paradoxical subreg forming a vector
29349         mode to a vec_concat.
29351 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29353         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
29354         -march=armv8.3-a variants.
29355         * config/arm/t-multilib: Likewise.
29356         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
29358 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29360         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
29361         to generate rtl.
29362         (cceq_ior_compare_complement): Give it a name so I can use it, and
29363         change boolean_or_operator predicate to boolean_operator so it can
29364         be used to generate a crand.
29365         (eqne): New code iterator.
29366         (bd/bd_neg): New code_attrs.
29367         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
29368         a single define_insn.
29369         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
29370         decrement (bdnzt/bdnzf/bdzt/bdzf).
29371         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
29372         with the new names of the branch decrement patterns, and added the
29373         names of the branch decrement conditional patterns.
29375 2018-01-08  Richard Biener  <rguenther@suse.de>
29377         PR tree-optimization/83563
29378         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
29379         cache.
29381 2018-01-08  Richard Biener  <rguenther@suse.de>
29383         PR middle-end/83713
29384         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
29386 2018-01-08  Richard Biener  <rguenther@suse.de>
29388         PR tree-optimization/83685
29389         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
29390         references to abnormals.
29392 2018-01-08  Richard Biener  <rguenther@suse.de>
29394         PR lto/83719
29395         * dwarf2out.c (output_indirect_strings): Handle empty
29396         skeleton_debug_str_hash.
29397         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
29399 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
29401         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
29402         (emit_store_direct): Likewise.
29403         (arc_trampoline_adjust_address): Likewise.
29404         (arc_asm_trampoline_template): New function.
29405         (arc_initialize_trampoline): Use asm_trampoline_template.
29406         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
29407         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
29408         * config/arc/arc.md (flush_icache): Delete pattern.
29410 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
29412         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
29413         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
29414         munaligned-access.
29416 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
29418         PR target/83681
29419         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
29420         by not USED_FOR_TARGET.
29421         (make_pass_resolve_sw_modes): Likewise.
29423 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
29425         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
29426         USED_FOR_TARGET.
29428 2018-01-08  Richard Biener  <rguenther@suse.de>
29430         PR middle-end/83580
29431         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
29433 2018-01-08  Richard Biener  <rguenther@suse.de>
29435         PR middle-end/83517
29436         * match.pd ((t * 2) / 2) -> t): Add missing :c.
29438 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
29440         PR middle-end/81897
29441         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
29442         basic blocks with a small number of successors.
29443         (convert_control_dep_chain_into_preds): Improve handling of
29444         forwarder blocks.
29445         (dump_predicates): Split apart into...
29446         (dump_pred_chain): ...here...
29447         (dump_pred_info): ...and here.
29448         (can_one_predicate_be_invalidated_p): Add debugging printfs.
29449         (can_chain_union_be_invalidated_p): Improve check for invalidation
29450         of paths.
29451         (uninit_uses_cannot_happen): Avoid unnecessary if
29452         convert_control_dep_chain_into_preds yielded nothing.
29454 2018-01-06  Martin Sebor  <msebor@redhat.com>
29456         PR tree-optimization/83640
29457         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
29458         subtracting negative offset from size.
29459         (builtin_access::overlap): Adjust offset bounds of the access to fall
29460         within the size of the object if possible.
29462 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
29464         PR rtl-optimization/83699
29465         * expmed.c (extract_bit_field_1): Restrict the vector usage of
29466         extract_bit_field_as_subreg to cases in which the extracted
29467         value is also a vector.
29469         * lra-constraints.c (process_alt_operands): Test for the equivalence
29470         substitutions when detecting a possible reload cycle.
29472 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
29474         PR debug/83480
29475         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
29476         by default if flag_selective_schedling{,2}.  Formatting fixes.
29478         PR rtl-optimization/83682
29479         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
29480         if it has non-VECTOR_MODE element mode.
29481         (vec_duplicate_p): Likewise.
29483         PR middle-end/83694
29484         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
29485         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
29487 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
29489         PR target/83604
29490         * config/i386/i386-builtin.def
29491         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
29492         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
29493         Require also OPTION_MASK_ISA_AVX512F in addition to
29494         OPTION_MASK_ISA_GFNI.
29495         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
29496         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
29497         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
29498         to OPTION_MASK_ISA_GFNI.
29499         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
29500         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
29501         OPTION_MASK_ISA_AVX512BW.
29502         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
29503         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
29504         addition to OPTION_MASK_ISA_GFNI.
29505         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
29506         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
29507         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
29508         to OPTION_MASK_ISA_GFNI.
29509         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
29510         a requirement for all ISAs rather than any of them with a few
29511         exceptions.
29512         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
29513         processing.
29514         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
29515         bitmasks to be enabled with 3 exceptions, instead of requiring any
29516         enabled ISA with lots of exceptions.
29517         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
29518         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
29519         Change avx512bw in isa attribute to avx512f.
29520         * config/i386/sgxintrin.h: Add license boilerplate.
29521         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
29522         to __AVX512F__ and __AVX512VL to __AVX512VL__.
29523         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
29524         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
29525         defined.
29526         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
29527         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
29528         temporarily sse2 rather than sse if not enabled already.
29530         PR target/83604
29531         * config/i386/sse.md (VI248_VLBW): Rename to ...
29532         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
29533         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
29534         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
29535         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
29536         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
29537         mode iterator instead of VI248_VLBW.
29539 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
29541         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
29542         (record_modified): Skip clobbers; add debug output.
29543         (param_change_prob): Use sreal frequencies.
29545 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
29547         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
29548         punt for user-aligned variables.
29550 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
29552         * tree-chrec.c (chrec_contains_symbols): Return true for
29553         POLY_INT_CST.
29555 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
29557         PR target/82439
29558         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
29559         of (x|y) == x for BICS pattern.
29561 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
29563         PR tree-optimization/83605
29564         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
29565         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
29566         can throw.
29568 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
29570         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
29571         * config/epiphany/rtems.h: New file.
29573 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
29574             Uros Bizjak  <ubizjak@gmail.com>
29576         PR target/83554
29577         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
29578         QIreg_operand instead of register_operand predicate.
29579         * config/i386/i386.c (ix86_rop_should_change_byte_p,
29580         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
29581         comments instead of -fmitigate[-_]rop.
29583 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29585         PR bootstrap/81926
29586         * cgraphunit.c (symbol_table::compile): Switch to text_section
29587         before calling assembly_start debug hook.
29588         * run-rtl-passes.c (run_rtl_passes): Likewise.
29589         Include output.h.
29591 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29593         * tree-vrp.c (extract_range_from_binary_expr_1): Check
29594         range_int_cst_p rather than !symbolic_range_p before calling
29595         extract_range_from_multiplicative_op_1.
29597 2018-01-04  Jeff Law  <law@redhat.com>
29599         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
29600         redundant test in assertion.
29602 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29604         * doc/rtl.texi: Document machine_mode wrapper classes.
29606 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29608         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
29609         using tree_to_uhwi.
29611 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29613         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
29614         the VEC_PERM_EXPR fold to fail.
29616 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
29618         PR debug/83585
29619         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
29620         to switched_sections.
29622 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29624         PR target/83680
29625         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
29626         test for d.testing.
29628 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
29630         PR target/83387
29631         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
29632         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
29634 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
29636         PR debug/83666
29637         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
29638         is BLKmode and bitpos not zero or mode change is needed.
29640 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
29642         PR target/83675
29643         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
29644         TARGET_VIS2.
29646 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
29648         PR target/83628
29649         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
29650         instead of MULT rtx.  Update all corresponding splitters.
29651         (*saddl_se): Ditto.
29652         (*ssub<modesuffix>): Ditto.
29653         (*ssubl_se): Ditto.
29654         (*cmp_sadd_di): Update split patterns.
29655         (*cmp_sadd_si): Ditto.
29656         (*cmp_sadd_sidi): Ditto.
29657         (*cmp_ssub_di): Ditto.
29658         (*cmp_ssub_si): Ditto.
29659         (*cmp_ssub_sidi): Ditto.
29660         * config/alpha/predicates.md (const23_operand): New predicate.
29661         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
29662         Look for ASHIFT, not MULT inner operand.
29663         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
29665 2018-01-04  Martin Liska  <mliska@suse.cz>
29667         PR gcov-profile/83669
29668         * gcov.c (output_intermediate_file): Add version to intermediate
29669         gcov file.
29670         * doc/gcov.texi: Document new field 'version' in intermediate
29671         file format. Fix location of '-k' option of gcov command.
29673 2018-01-04  Martin Liska  <mliska@suse.cz>
29675         PR ipa/82352
29676         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
29678 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
29680         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
29682 2018-01-03  Martin Sebor  <msebor@redhat.com>
29684         PR tree-optimization/83655
29685         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
29686         checking calls with invalid arguments.
29688 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29690         * tree-vect-stmts.c (vect_get_store_rhs): New function.
29691         (vectorizable_mask_load_store): Delete.
29692         (vectorizable_call): Return false for masked loads and stores.
29693         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
29694         instead of gimple_assign_rhs1.
29695         (vectorizable_load): Handle IFN_MASK_LOAD.
29696         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
29698 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29700         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
29701         split out from..,
29702         (vectorizable_mask_load_store): ...here.
29703         (vectorizable_load): ...and here.
29705 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29707         * tree-vect-stmts.c (vect_build_all_ones_mask)
29708         (vect_build_zero_merge_argument): New functions, split out from...
29709         (vectorizable_load): ...here.
29711 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29713         * tree-vect-stmts.c (vect_check_store_rhs): New function,
29714         split out from...
29715         (vectorizable_mask_load_store): ...here.
29716         (vectorizable_store): ...and here.
29718 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29720         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
29721         split out from...
29722         (vectorizable_mask_load_store): ...here.
29724 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29726         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
29727         (vect_model_store_cost): Take a vec_load_store_type instead of a
29728         vect_def_type.
29729         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
29730         (vect_model_store_cost): Take a vec_load_store_type instead of a
29731         vect_def_type.
29732         (vectorizable_mask_load_store): Update accordingly.
29733         (vectorizable_store): Likewise.
29734         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
29736 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29738         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
29739         IFN_MASK_LOAD calls here rather than...
29740         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
29742 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29743             Alan Hayward  <alan.hayward@arm.com>
29744             David Sherwood  <david.sherwood@arm.com>
29746         * expmed.c (extract_bit_field_1): For vector extracts,
29747         fall back to extract_bit_field_as_subreg if vec_extract
29748         isn't available.
29750 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29751             Alan Hayward  <alan.hayward@arm.com>
29752             David Sherwood  <david.sherwood@arm.com>
29754         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
29755         they are variable or constant sized.
29756         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
29757         slots for constant-sized data.
29759 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29760             Alan Hayward  <alan.hayward@arm.com>
29761             David Sherwood  <david.sherwood@arm.com>
29763         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
29764         handling COND_EXPRs with boolean comparisons, try to find a better
29765         basis for the mask type than the boolean itself.
29767 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29769         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
29770         is calculated and how it can be overridden.
29771         * genmodes.c (max_bitsize_mode_any_mode): New variable.
29772         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
29773         if defined.
29774         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
29775         if nonzero.
29777 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29778             Alan Hayward  <alan.hayward@arm.com>
29779             David Sherwood  <david.sherwood@arm.com>
29781         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
29782         Remove the mode argument.
29783         (aarch64_simd_valid_immediate): Remove the mode and inverse
29784         arguments.
29785         * config/aarch64/iterators.md (bitsize): New iterator.
29786         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
29787         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
29788         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
29789         aarch64_simd_valid_immediate.
29790         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
29791         (aarch64_reg_or_bic_imm): Likewise.
29792         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
29793         with an insn_type enum and msl with a modifier_type enum.
29794         Replace element_width with a scalar_mode.  Change the shift
29795         to unsigned int.  Add constructors for scalar_float_mode and
29796         scalar_int_mode elements.
29797         (aarch64_vect_float_const_representable_p): Delete.
29798         (aarch64_can_const_movi_rtx_p)
29799         (aarch64_simd_scalar_immediate_valid_for_move)
29800         (aarch64_simd_make_constant): Update call to
29801         aarch64_simd_valid_immediate.
29802         (aarch64_advsimd_valid_immediate_hs): New function.
29803         (aarch64_advsimd_valid_immediate): Likewise.
29804         (aarch64_simd_valid_immediate): Remove mode and inverse
29805         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
29806         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
29807         and aarch64_float_const_representable_p on the result.
29808         (aarch64_output_simd_mov_immediate): Remove mode argument.
29809         Update call to aarch64_simd_valid_immediate and use of
29810         simd_immediate_info.
29811         (aarch64_output_scalar_simd_mov_immediate): Update call
29812         accordingly.
29814 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29815             Alan Hayward  <alan.hayward@arm.com>
29816             David Sherwood  <david.sherwood@arm.com>
29818         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
29819         (mode_nunits): Likewise CONST_MODE_NUNITS.
29820         * machmode.def (ADJUST_NUNITS): Document.
29821         * genmodes.c (mode_data::need_nunits_adj): New field.
29822         (blank_mode): Update accordingly.
29823         (adj_nunits): New variable.
29824         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
29825         parameter.
29826         (emit_mode_size_inline): Set need_bytesize_adj for all modes
29827         listed in adj_nunits.
29828         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
29829         listed in adj_nunits.  Don't emit case statements for such modes.
29830         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
29831         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
29832         nothing if adj_nunits is nonnull.
29833         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
29834         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
29835         (emit_mode_fbit): Update use of print_maybe_const_decl.
29836         (emit_move_size): Likewise.  Treat the array as non-const
29837         if adj_nunits.
29838         (emit_mode_adjustments): Handle adj_nunits.
29840 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29842         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
29843         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
29844         (VECTOR_MODES): Use it.
29845         (make_vector_modes): Take the prefix as an argument.
29847 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29848             Alan Hayward  <alan.hayward@arm.com>
29849             David Sherwood  <david.sherwood@arm.com>
29851         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
29852         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
29853         for MODE_VECTOR_BOOL.
29854         * machmode.def (VECTOR_BOOL_MODE): Document.
29855         * genmodes.c (VECTOR_BOOL_MODE): New macro.
29856         (make_vector_bool_mode): New function.
29857         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
29858         MODE_VECTOR_BOOL.
29859         * lto-streamer-in.c (lto_input_mode_table): Likewise.
29860         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
29861         Likewise.
29862         * stor-layout.c (int_mode_for_mode): Likewise.
29863         * tree.c (build_vector_type_for_mode): Likewise.
29864         * varasm.c (output_constant_pool_2): Likewise.
29865         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
29866         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
29867         for MODE_VECTOR_BOOL.
29868         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
29869         of mode class checks.
29870         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
29871         instead of a list of mode class checks.
29872         (expand_vector_scalar_condition): Likewise.
29873         (type_for_widest_vector_mode): Handle BImode as an inner mode.
29875 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29876             Alan Hayward  <alan.hayward@arm.com>
29877             David Sherwood  <david.sherwood@arm.com>
29879         * machmode.h (mode_size): Change from unsigned short to
29880         poly_uint16_pod.
29881         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
29882         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
29883         or if measurement_type is not polynomial.
29884         (fixed_size_mode::includes_p): Check for constant-sized modes.
29885         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
29886         return a poly_uint16 rather than an unsigned short.
29887         (emit_mode_size): Change the type of mode_size from unsigned short
29888         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
29889         (emit_mode_adjustments): Cope with polynomial vector sizes.
29890         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
29891         for GET_MODE_SIZE.
29892         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
29893         for GET_MODE_SIZE.
29894         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
29895         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
29896         * caller-save.c (setup_save_areas): Likewise.
29897         (replace_reg_with_saved_mem): Likewise.
29898         * calls.c (emit_library_call_value_1): Likewise.
29899         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
29900         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
29901         (gen_lowpart_for_combine): Likewise.
29902         * convert.c (convert_to_integer_1): Likewise.
29903         * cse.c (equiv_constant, cse_insn): Likewise.
29904         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
29905         (cselib_subst_to_values): Likewise.
29906         * dce.c (word_dce_process_block): Likewise.
29907         * df-problems.c (df_word_lr_mark_ref): Likewise.
29908         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
29909         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
29910         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
29911         (rtl_for_decl_location): Likewise.
29912         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
29913         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
29914         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
29915         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
29916         (expand_expr_real_1): Likewise.
29917         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
29918         (pad_below): Likewise.
29919         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
29920         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
29921         * ira.c (get_subreg_tracking_sizes): Likewise.
29922         * ira-build.c (ira_create_allocno_objects): Likewise.
29923         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
29924         (ira_sort_regnos_for_alter_reg): Likewise.
29925         * ira-costs.c (record_operand_costs): Likewise.
29926         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
29927         (resolve_simple_move): Likewise.
29928         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
29929         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
29930         (lra_constraints): Likewise.
29931         (CONST_POOL_OK_P): Reject variable-sized modes.
29932         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
29933         (add_pseudo_to_slot, lra_spill): Likewise.
29934         * omp-low.c (omp_clause_aligned_alignment): Likewise.
29935         * optabs-query.c (get_best_extraction_insn): Likewise.
29936         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
29937         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
29938         (expand_mult_highpart, valid_multiword_target_p): Likewise.
29939         * recog.c (offsettable_address_addr_space_p): Likewise.
29940         * regcprop.c (maybe_mode_change): Likewise.
29941         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
29942         * regrename.c (build_def_use): Likewise.
29943         * regstat.c (dump_reg_info): Likewise.
29944         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
29945         (find_reloads, find_reloads_subreg_address): Likewise.
29946         * reload1.c (eliminate_regs_1): Likewise.
29947         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
29948         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
29949         (simplify_binary_operation_1, simplify_subreg): Likewise.
29950         * targhooks.c (default_function_arg_padding): Likewise.
29951         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
29952         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
29953         (verify_gimple_assign_ternary): Likewise.
29954         * tree-inline.c (estimate_move_cost): Likewise.
29955         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29956         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
29957         (get_address_cost_ainc): Likewise.
29958         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
29959         (vect_supportable_dr_alignment): Likewise.
29960         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
29961         (vectorizable_reduction): Likewise.
29962         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
29963         (vectorizable_operation, vectorizable_load): Likewise.
29964         * tree.c (build_same_sized_truth_vector_type): Likewise.
29965         * valtrack.c (cleanup_auto_inc_dec): Likewise.
29966         * var-tracking.c (emit_note_insn_var_location): Likewise.
29967         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
29968         (ADDR_VEC_ALIGN): Likewise.
29970 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29971             Alan Hayward  <alan.hayward@arm.com>
29972             David Sherwood  <david.sherwood@arm.com>
29974         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
29975         unsigned short.
29976         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
29977         or if measurement_type is polynomial.
29978         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
29979         * combine.c (make_extraction): Likewise.
29980         * dse.c (find_shift_sequence): Likewise.
29981         * dwarf2out.c (mem_loc_descriptor): Likewise.
29982         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
29983         (extract_bit_field, extract_low_bits): Likewise.
29984         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
29985         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
29986         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
29987         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
29988         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
29989         * reload.c (find_reloads): Likewise.
29990         * reload1.c (alter_reg): Likewise.
29991         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
29992         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
29993         * tree-if-conv.c (predicate_mem_writes): Likewise.
29994         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
29995         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
29996         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
29997         * valtrack.c (dead_debug_insert_temp): Likewise.
29998         * varasm.c (mergeable_constant_section): Likewise.
29999         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
30001 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30002             Alan Hayward  <alan.hayward@arm.com>
30003             David Sherwood  <david.sherwood@arm.com>
30005         * expr.c (expand_assignment): Cope with polynomial mode sizes
30006         when assigning to a CONCAT.
30008 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30009             Alan Hayward  <alan.hayward@arm.com>
30010             David Sherwood  <david.sherwood@arm.com>
30012         * machmode.h (mode_precision): Change from unsigned short to
30013         poly_uint16_pod.
30014         (mode_to_precision): Return a poly_uint16 rather than an unsigned
30015         short.
30016         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
30017         or if measurement_type is not polynomial.
30018         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
30019         in which the mode is already known to be a scalar_int_mode.
30020         * genmodes.c (emit_mode_precision): Change the type of mode_precision
30021         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
30022         initializer.
30023         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30024         for GET_MODE_PRECISION.
30025         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30026         for GET_MODE_PRECISION.
30027         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
30028         as polynomial.
30029         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
30030         (expand_field_assignment, make_extraction): Likewise.
30031         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
30032         (get_last_value): Likewise.
30033         * convert.c (convert_to_integer_1): Likewise.
30034         * cse.c (cse_insn): Likewise.
30035         * expr.c (expand_expr_real_1): Likewise.
30036         * lra-constraints.c (simplify_operand_subreg): Likewise.
30037         * optabs-query.c (can_atomic_load_p): Likewise.
30038         * optabs.c (expand_atomic_load): Likewise.
30039         (expand_atomic_store): Likewise.
30040         * ree.c (combine_reaching_defs): Likewise.
30041         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
30042         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
30043         * tree.h (type_has_mode_precision_p): Likewise.
30044         * ubsan.c (instrument_si_overflow): Likewise.
30046 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30047             Alan Hayward  <alan.hayward@arm.com>
30048             David Sherwood  <david.sherwood@arm.com>
30050         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
30051         polynomial numbers of units.
30052         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
30053         (valid_vector_subparts_p): New function.
30054         (build_vector_type): Remove temporary shim and take the number
30055         of units as a poly_uint64 rather than an int.
30056         (build_opaque_vector_type): Take the number of units as a
30057         poly_uint64 rather than an int.
30058         * tree.c (build_vector_from_ctor): Handle polynomial
30059         TYPE_VECTOR_SUBPARTS.
30060         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
30061         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
30062         (build_vector_from_val): If the number of units is variable,
30063         use build_vec_duplicate_cst for constant operands and
30064         VEC_DUPLICATE_EXPR otherwise.
30065         (make_vector_type): Remove temporary is_constant ().
30066         (build_vector_type, build_opaque_vector_type): Take the number of
30067         units as a poly_uint64 rather than an int.
30068         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
30069         VECTOR_CST_NELTS.
30070         * cfgexpand.c (expand_debug_expr): Likewise.
30071         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
30072         (store_constructor, expand_expr_real_1): Likewise.
30073         (const_scalar_mask_from_tree): Likewise.
30074         * fold-const-call.c (fold_const_reduction): Likewise.
30075         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
30076         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
30077         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
30078         (fold_relational_const): Likewise.
30079         (native_interpret_vector): Likewise.  Change the size from an
30080         int to an unsigned int.
30081         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
30082         TYPE_VECTOR_SUBPARTS.
30083         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
30084         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
30085         duplicating a non-constant operand into a variable-length vector.
30086         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
30087         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
30088         * ipa-icf.c (sem_variable::equals): Likewise.
30089         * match.pd: Likewise.
30090         * omp-simd-clone.c (simd_clone_subparts): Likewise.
30091         * print-tree.c (print_node): Likewise.
30092         * stor-layout.c (layout_type): Likewise.
30093         * targhooks.c (default_builtin_vectorization_cost): Likewise.
30094         * tree-cfg.c (verify_gimple_comparison): Likewise.
30095         (verify_gimple_assign_binary): Likewise.
30096         (verify_gimple_assign_ternary): Likewise.
30097         (verify_gimple_assign_single): Likewise.
30098         * tree-pretty-print.c (dump_generic_node): Likewise.
30099         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30100         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
30101         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
30102         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
30103         (vect_shift_permute_load_chain): Likewise.
30104         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
30105         (expand_vector_condition, optimize_vector_constructor): Likewise.
30106         (lower_vec_perm, get_compute_type): Likewise.
30107         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
30108         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
30109         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
30110         (vect_recog_mask_conversion_pattern): Likewise.
30111         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
30112         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
30113         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30114         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
30115         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
30116         (vectorizable_shift, vectorizable_operation, vectorizable_store)
30117         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
30118         (supportable_widening_operation): Likewise.
30119         (supportable_narrowing_operation): Likewise.
30120         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
30121         Likewise.
30122         * varasm.c (output_constant): Likewise.
30124 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30125             Alan Hayward  <alan.hayward@arm.com>
30126             David Sherwood  <david.sherwood@arm.com>
30128         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
30129         so that both the length == 3 and length != 3 cases set up their
30130         own permute vectors.  Add comments explaining why we know the
30131         number of elements is constant.
30132         (vect_permute_load_chain): Likewise.
30134 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30135             Alan Hayward  <alan.hayward@arm.com>
30136             David Sherwood  <david.sherwood@arm.com>
30138         * machmode.h (mode_nunits): Change from unsigned char to
30139         poly_uint16_pod.
30140         (ONLY_FIXED_SIZE_MODES): New macro.
30141         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
30142         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
30143         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
30144         New typedefs.
30145         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
30146         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
30147         or if measurement_type is not polynomial.
30148         * genmodes.c (ZERO_COEFFS): New macro.
30149         (emit_mode_nunits_inline): Make mode_nunits_inline return a
30150         poly_uint16.
30151         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
30152         Use ZERO_COEFFS when emitting initializers.
30153         * data-streamer.h (bp_pack_poly_value): New function.
30154         (bp_unpack_poly_value): Likewise.
30155         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30156         for GET_MODE_NUNITS.
30157         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30158         for GET_MODE_NUNITS.
30159         * tree.c (make_vector_type): Remove temporary shim and make
30160         the real function take the number of units as a poly_uint64
30161         rather than an int.
30162         (build_vector_type_for_mode): Handle polynomial nunits.
30163         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
30164         * emit-rtl.c (const_vec_series_p_1): Likewise.
30165         (gen_rtx_CONST_VECTOR): Likewise.
30166         * fold-const.c (test_vec_duplicate_folding): Likewise.
30167         * genrecog.c (validate_pattern): Likewise.
30168         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
30169         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
30170         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
30171         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
30172         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
30173         * rtlanal.c (subreg_get_info): Likewise.
30174         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
30175         (vect_grouped_load_supported): Likewise.
30176         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
30177         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
30178         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
30179         (simplify_const_unary_operation, simplify_binary_operation_1)
30180         (simplify_const_binary_operation, simplify_ternary_operation)
30181         (test_vector_ops_duplicate, test_vector_ops): Likewise.
30182         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
30183         instead of CONST_VECTOR_NUNITS.
30184         * varasm.c (output_constant_pool_2): Likewise.
30185         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
30186         explicit-encoded elements in the XVEC for variable-length vectors.
30188 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30190         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
30192 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30193             Alan Hayward  <alan.hayward@arm.com>
30194             David Sherwood  <david.sherwood@arm.com>
30196         * coretypes.h (fixed_size_mode): Declare.
30197         (fixed_size_mode_pod): New typedef.
30198         * builtins.h (target_builtins::x_apply_args_mode)
30199         (target_builtins::x_apply_result_mode): Change type to
30200         fixed_size_mode_pod.
30201         * builtins.c (apply_args_size, apply_result_size, result_vector)
30202         (expand_builtin_apply_args_1, expand_builtin_apply)
30203         (expand_builtin_return): Update accordingly.
30205 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30207         * cse.c (hash_rtx_cb): Hash only the encoded elements.
30208         * cselib.c (cselib_hash_rtx): Likewise.
30209         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
30210         CONST_VECTOR encoding.
30212 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
30213             Jeff Law  <law@redhat.com>
30215         PR target/83641
30216         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
30217         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
30218         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
30219         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
30221         PR target/83641
30222         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
30223         explicitly probe *sp in a noreturn function if there were any callee
30224         register saves or frame pointer is needed.
30226 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
30228         PR debug/83621
30229         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
30230         BLKmode for ternary, binary or unary expressions.
30232         PR debug/83645
30233         * var-tracking.c (delete_vta_debug_insn): New inline function.
30234         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
30235         insns from get_insns () to NULL instead of each bb separately.
30236         Use delete_vta_debug_insn.  No longer static.
30237         (vt_debug_insns_local, variable_tracking_main_1): Adjust
30238         delete_vta_debug_insns callers.
30239         * rtl.h (delete_vta_debug_insns): Declare.
30240         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
30241         instead of variable_tracking_main.
30243 2018-01-03  Martin Sebor  <msebor@redhat.com>
30245         PR tree-optimization/83603
30246         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
30247         arguments past the endof the argument list in functions declared
30248         without a prototype.
30249         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
30250         Avoid checking when arguments are null.
30252 2018-01-03  Martin Sebor  <msebor@redhat.com>
30254         PR c/83559
30255         * doc/extend.texi (attribute const): Fix a typo.
30256         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
30257         issuing -Wsuggest-attribute for void functions.
30259 2018-01-03  Martin Sebor  <msebor@redhat.com>
30261         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
30262         offset_int::from instead of wide_int::to_shwi.
30263         (maybe_diag_overlap): Remove assertion.
30264         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
30265         * gimple-ssa-sprintf.c (format_directive): Same.
30266         (parse_directive): Same.
30267         (sprintf_dom_walker::compute_format_length): Same.
30268         (try_substitute_return_value): Same.
30270 2018-01-03  Jeff Law  <law@redhat.com>
30272         PR middle-end/83654
30273         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
30274         non-constant residual for zero at runtime and avoid probing in
30275         that case.  Reorganize code for trailing problem to mirror handling
30276         of the residual.
30278 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
30280         PR tree-optimization/83501
30281         * tree-ssa-strlen.c (get_string_cst): New.
30282         (handle_char_store): Call get_string_cst.
30284 2018-01-03  Martin Liska  <mliska@suse.cz>
30286         PR tree-optimization/83593
30287         * tree-ssa-strlen.c: Include tree-cfg.h.
30288         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
30289         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
30290         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
30291         to false.
30292         (strlen_dom_walker::before_dom_children): Call
30293         gimple_purge_dead_eh_edges. Dump tranformation with details
30294         dump flags.
30295         (strlen_dom_walker::before_dom_children): Update call by adding
30296         new argument cleanup_eh.
30297         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
30299 2018-01-03  Martin Liska  <mliska@suse.cz>
30301         PR ipa/83549
30302         * cif-code.def (VARIADIC_THUNK): New enum value.
30303         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
30304         thunks.
30306 2018-01-03  Jan Beulich  <jbeulich@suse.com>
30308         * sse.md (mov<mode>_internal): Tighten condition for when to use
30309         vmovdqu<ssescalarsize> for TI and OI modes.
30311 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
30313         Update copyright years.
30315 2018-01-03  Martin Liska  <mliska@suse.cz>
30317         PR ipa/83594
30318         * ipa-visibility.c (function_and_variable_visibility): Skip
30319         functions with noipa attribure.
30321 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
30323         * gcc.c (process_command): Update copyright notice dates.
30324         * gcov-dump.c (print_version): Ditto.
30325         * gcov.c (print_version): Ditto.
30326         * gcov-tool.c (print_version): Ditto.
30327         * gengtype.c (create_file): Ditto.
30328         * doc/cpp.texi: Bump @copying's copyright year.
30329         * doc/cppinternals.texi: Ditto.
30330         * doc/gcc.texi: Ditto.
30331         * doc/gccint.texi: Ditto.
30332         * doc/gcov.texi: Ditto.
30333         * doc/install.texi: Ditto.
30334         * doc/invoke.texi: Ditto.
30336 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30338         * vector-builder.h (vector_builder::m_full_nelts): Change from
30339         unsigned int to poly_uint64.
30340         (vector_builder::full_nelts): Update prototype accordingly.
30341         (vector_builder::new_vector): Likewise.
30342         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
30343         (vector_builder::operator ==): Likewise.
30344         (vector_builder::finalize): Likewise.
30345         * int-vector-builder.h (int_vector_builder::int_vector_builder):
30346         Take the number of elements as a poly_uint64 rather than an
30347         unsigned int.
30348         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
30349         from unsigned int to poly_uint64.
30350         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
30351         (vec_perm_indices::new_vector): Likewise.
30352         (vec_perm_indices::length): Likewise.
30353         (vec_perm_indices::nelts_per_input): Likewise.
30354         (vec_perm_indices::input_nelts): Likewise.
30355         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
30356         number of elements per input as a poly_uint64 rather than an
30357         unsigned int.  Use the original encoding for variable-length
30358         vectors, rather than clamping each individual element.
30359         For the second and subsequent elements in each pattern,
30360         clamp the step and base before clamping their sum.
30361         (vec_perm_indices::series_p): Handle polynomial element counts.
30362         (vec_perm_indices::all_in_range_p): Likewise.
30363         (vec_perm_indices_to_tree): Likewise.
30364         (vec_perm_indices_to_rtx): Likewise.
30365         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
30366         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
30367         (tree_vector_builder::new_binary_operation): Handle polynomial
30368         element counts.  Return false if we need to know the number
30369         of elements at compile time.
30370         * fold-const.c (fold_vec_perm): Punt if the number of elements
30371         isn't known at compile time.
30373 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30375         * vec-perm-indices.h (vec_perm_builder): Change element type
30376         from HOST_WIDE_INT to poly_int64.
30377         (vec_perm_indices::element_type): Update accordingly.
30378         (vec_perm_indices::clamp): Handle polynomial element_types.
30379         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
30380         (vec_perm_indices::all_in_range_p): Likewise.
30381         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
30382         than shwi trees.
30383         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
30384         polynomial vec_perm_indices element types.
30385         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
30386         * fold-const.c (fold_vec_perm): Likewise.
30387         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
30388         * tree-vect-generic.c (lower_vec_perm): Likewise.
30389         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
30390         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
30391         element type to HOST_WIDE_INT.
30393 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30394             Alan Hayward  <alan.hayward@arm.com>
30395             David Sherwood  <david.sherwood@arm.com>
30397         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
30398         rather than an int.  Use plus_constant.
30399         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
30400         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
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         * calls.c (emit_call_1, expand_call): Change struct_value_size from
30407         a HOST_WIDE_INT to a poly_int64.
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         * calls.c (load_register_parameters): Cope with polynomial
30414         mode sizes.  Require a constant size for BLKmode parameters
30415         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
30416         forces a parameter to be padded at the lsb end in order to
30417         fill a complete number of words, require the parameter size
30418         to be ordered wrt UNITS_PER_WORD.
30420 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30421             Alan Hayward  <alan.hayward@arm.com>
30422             David Sherwood  <david.sherwood@arm.com>
30424         * reload1.c (spill_stack_slot_width): Change element type
30425         from unsigned int to poly_uint64_pod.
30426         (alter_reg): Treat mode sizes as polynomial.
30428 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30429             Alan Hayward  <alan.hayward@arm.com>
30430             David Sherwood  <david.sherwood@arm.com>
30432         * reload.c (complex_word_subreg_p): New function.
30433         (reload_inner_reg_of_subreg, push_reload): Use it.
30435 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30436             Alan Hayward  <alan.hayward@arm.com>
30437             David Sherwood  <david.sherwood@arm.com>
30439         * lra-constraints.c (process_alt_operands): Reject matched
30440         operands whose sizes aren't ordered.
30441         (match_reload): Refer to this check here.
30443 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30444             Alan Hayward  <alan.hayward@arm.com>
30445             David Sherwood  <david.sherwood@arm.com>
30447         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
30448         that the mode size is in the set {1, 2, 4, 8, 16}.
30450 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30451             Alan Hayward  <alan.hayward@arm.com>
30452             David Sherwood  <david.sherwood@arm.com>
30454         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
30455         Use plus_constant instead of gen_rtx_PLUS.
30457 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30458             Alan Hayward  <alan.hayward@arm.com>
30459             David Sherwood  <david.sherwood@arm.com>
30461         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
30462         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
30463         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
30464         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
30465         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
30466         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
30467         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
30468         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
30469         * config/i386/i386.c (ix86_push_rounding): ...this new function.
30470         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
30471         a poly_int64.
30472         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
30473         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
30474         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
30475         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
30476         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
30477         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
30478         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
30479         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
30480         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
30481         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
30482         function.
30483         * expr.c (emit_move_resolve_push): Treat the input and result
30484         of PUSH_ROUNDING as a poly_int64.
30485         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
30486         (emit_push_insn): Likewise.
30487         * lra-eliminations.c (mark_not_eliminable): Likewise.
30488         * recog.c (push_operand): Likewise.
30489         * reload1.c (elimination_effects): Likewise.
30490         * rtlanal.c (nonzero_bits1): Likewise.
30491         * calls.c (store_one_arg): Likewise.  Require the padding to be
30492         known at compile time.
30494 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30495             Alan Hayward  <alan.hayward@arm.com>
30496             David Sherwood  <david.sherwood@arm.com>
30498         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
30499         Use plus_constant instead of gen_rtx_PLUS.
30501 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30502             Alan Hayward  <alan.hayward@arm.com>
30503             David Sherwood  <david.sherwood@arm.com>
30505         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
30506         rather than an int.
30508 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30509             Alan Hayward  <alan.hayward@arm.com>
30510             David Sherwood  <david.sherwood@arm.com>
30512         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
30513         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
30514         via stack temporaries.  Treat the mode size as polynomial too.
30516 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30517             Alan Hayward  <alan.hayward@arm.com>
30518             David Sherwood  <david.sherwood@arm.com>
30520         * expr.c (expand_expr_real_2): When handling conversions involving
30521         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
30522         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
30523         as a poly_uint64 too.
30525 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30526             Alan Hayward  <alan.hayward@arm.com>
30527             David Sherwood  <david.sherwood@arm.com>
30529         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
30531 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30532             Alan Hayward  <alan.hayward@arm.com>
30533             David Sherwood  <david.sherwood@arm.com>
30535         * combine.c (can_change_dest_mode): Handle polynomial
30536         REGMODE_NATURAL_SIZE.
30537         * expmed.c (store_bit_field_1): Likewise.
30538         * expr.c (store_constructor): Likewise.
30539         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
30540         and polynomial REGMODE_NATURAL_SIZE.
30541         (gen_lowpart_common): Likewise.
30542         * reginfo.c (record_subregs_of_mode): Likewise.
30543         * rtlanal.c (read_modify_subreg_p): Likewise.
30545 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30546             Alan Hayward  <alan.hayward@arm.com>
30547             David Sherwood  <david.sherwood@arm.com>
30549         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
30550         numbers of elements.
30552 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30553             Alan Hayward  <alan.hayward@arm.com>
30554             David Sherwood  <david.sherwood@arm.com>
30556         * match.pd: Cope with polynomial numbers of vector elements.
30558 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30559             Alan Hayward  <alan.hayward@arm.com>
30560             David Sherwood  <david.sherwood@arm.com>
30562         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
30563         in a POINTER_PLUS_EXPR.
30565 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30566             Alan Hayward  <alan.hayward@arm.com>
30567             David Sherwood  <david.sherwood@arm.com>
30569         * omp-simd-clone.c (simd_clone_subparts): New function.
30570         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
30571         (ipa_simd_modify_function_body): Likewise.
30573 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30574             Alan Hayward  <alan.hayward@arm.com>
30575             David Sherwood  <david.sherwood@arm.com>
30577         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
30578         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
30579         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
30580         (expand_vector_condition, vector_element): Likewise.
30581         (subparts_gt): New function.
30582         (get_compute_type): Use subparts_gt.
30583         (count_type_subparts): Delete.
30584         (expand_vector_operations_1): Use subparts_gt instead of
30585         count_type_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-data-refs.c (vect_no_alias_p): Replace with...
30592         (vect_compile_time_alias): ...this new function.  Do the calculation
30593         on poly_ints rather than trees.
30594         (vect_prune_runtime_alias_test_list): Update call accordingly.
30596 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30597             Alan Hayward  <alan.hayward@arm.com>
30598             David Sherwood  <david.sherwood@arm.com>
30600         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
30601         numbers of units.
30602         (vect_schedule_slp_instance): Likewise.
30604 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30605             Alan Hayward  <alan.hayward@arm.com>
30606             David Sherwood  <david.sherwood@arm.com>
30608         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
30609         constant and extern definitions for variable-length vectors.
30610         (vect_get_constant_vectors): Note that the number of units
30611         is known to be constant.
30613 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30614             Alan Hayward  <alan.hayward@arm.com>
30615             David Sherwood  <david.sherwood@arm.com>
30617         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
30618         of units as polynomial.  Choose between WIDE and NARROW based
30619         on multiple_p.
30621 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30622             Alan Hayward  <alan.hayward@arm.com>
30623             David Sherwood  <david.sherwood@arm.com>
30625         * tree-vect-stmts.c (simd_clone_subparts): New function.
30626         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
30628 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30629             Alan Hayward  <alan.hayward@arm.com>
30630             David Sherwood  <david.sherwood@arm.com>
30632         * tree-vect-stmts.c (vectorizable_call): Treat the number of
30633         vectors as polynomial.  Use build_index_vector for
30634         IFN_GOMP_SIMD_LANE.
30636 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30637             Alan Hayward  <alan.hayward@arm.com>
30638             David Sherwood  <david.sherwood@arm.com>
30640         * tree-vect-stmts.c (get_load_store_type): Treat the number of
30641         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
30642         for variable-length vectors.
30643         (vectorizable_mask_load_store): Treat the number of units as
30644         polynomial, asserting that it is constant if the condition has
30645         already been enforced.
30646         (vectorizable_store, vectorizable_load): Likewise.
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         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
30653         of units as polynomial.  Punt if we can't tell at compile time
30654         which vector contains the final result.
30656 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30657             Alan Hayward  <alan.hayward@arm.com>
30658             David Sherwood  <david.sherwood@arm.com>
30660         * tree-vect-loop.c (vectorizable_induction): Treat the number
30661         of units as polynomial.  Punt on SLP inductions.  Use an integer
30662         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
30663         cast of such a series for variable-length floating-point
30664         reductions.
30666 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30667             Alan Hayward  <alan.hayward@arm.com>
30668             David Sherwood  <david.sherwood@arm.com>
30670         * tree.h (build_index_vector): Declare.
30671         * tree.c (build_index_vector): New function.
30672         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
30673         of units as polynomial, forcibly converting it to a constant if
30674         vectorizable_reduction has already enforced the condition.
30675         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
30676         to create a {1,2,3,...} vector.
30677         (vectorizable_reduction): Treat the number of units as polynomial.
30678         Choose vectype_in based on the largest scalar element size rather
30679         than the smallest number of units.  Enforce the restrictions
30680         relied on above.
30682 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30683             Alan Hayward  <alan.hayward@arm.com>
30684             David Sherwood  <david.sherwood@arm.com>
30686         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
30687         number of units as polynomial.
30689 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30690             Alan Hayward  <alan.hayward@arm.com>
30691             David Sherwood  <david.sherwood@arm.com>
30693         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
30694         * target.def (autovectorize_vector_sizes): Return the vector sizes
30695         by pointer, using vector_sizes rather than a bitmask.
30696         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
30697         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
30698         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
30699         Likewise.
30700         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
30701         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
30702         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
30703         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
30704         * omp-general.c (omp_max_vf): Likewise.
30705         * omp-low.c (omp_clause_aligned_alignment): Likewise.
30706         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
30707         * tree-vect-loop.c (vect_analyze_loop): Likewise.
30708         * tree-vect-slp.c (vect_slp_bb): Likewise.
30709         * doc/tm.texi: Regenerate.
30710         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
30711         to a poly_uint64.
30712         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
30713         the vector size as a poly_uint64 rather than an unsigned int.
30714         (current_vector_size): Change from an unsigned int to a poly_uint64.
30715         (get_vectype_for_scalar_type): Update accordingly.
30716         * tree.h (build_truth_vector_type): Take the size and number of
30717         units as a poly_uint64 rather than an unsigned int.
30718         (build_vector_type): Add a temporary overload that takes
30719         the number of units as a poly_uint64 rather than an unsigned int.
30720         * tree.c (make_vector_type): Likewise.
30721         (build_truth_vector_type): Take the number of units as a poly_uint64
30722         rather than an unsigned int.
30724 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30725             Alan Hayward  <alan.hayward@arm.com>
30726             David Sherwood  <david.sherwood@arm.com>
30728         * target.def (get_mask_mode): Take the number of units and length
30729         as poly_uint64s rather than unsigned ints.
30730         * targhooks.h (default_get_mask_mode): Update accordingly.
30731         * targhooks.c (default_get_mask_mode): Likewise.
30732         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
30733         * doc/tm.texi: Regenerate.
30735 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30736             Alan Hayward  <alan.hayward@arm.com>
30737             David Sherwood  <david.sherwood@arm.com>
30739         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
30740         * omp-general.c (omp_max_vf): Likewise.
30741         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
30742         (expand_omp_simd): Handle polynomial safelen.
30743         * omp-low.c (omplow_simd_context): Add a default constructor.
30744         (omplow_simd_context::max_vf): Change from int to poly_uint64.
30745         (lower_rec_simd_input_clauses): Update accordingly.
30746         (lower_rec_input_clauses): Likewise.
30748 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30749             Alan Hayward  <alan.hayward@arm.com>
30750             David Sherwood  <david.sherwood@arm.com>
30752         * tree-vectorizer.h (vect_nunits_for_cost): New function.
30753         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
30754         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
30755         (vect_analyze_slp_cost): Likewise.
30756         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
30757         (vect_model_load_cost): Likewise.
30759 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30760             Alan Hayward  <alan.hayward@arm.com>
30761             David Sherwood  <david.sherwood@arm.com>
30763         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
30764         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
30765         from an unsigned int * to a poly_uint64_pod *.
30766         (calculate_unrolling_factor): New function.
30767         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
30769 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30770             Alan Hayward  <alan.hayward@arm.com>
30771             David Sherwood  <david.sherwood@arm.com>
30773         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
30774         from an unsigned int to a poly_uint64.
30775         (_loop_vec_info::slp_unrolling_factor): Likewise.
30776         (_loop_vec_info::vectorization_factor): Change from an int
30777         to a poly_uint64.
30778         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
30779         (vect_get_num_vectors): New function.
30780         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
30781         (vect_get_num_copies): Use vect_get_num_vectors.
30782         (vect_analyze_data_ref_dependences): Change max_vf from an int *
30783         to an unsigned int *.
30784         (vect_analyze_data_refs): Change min_vf from an int * to a
30785         poly_uint64 *.
30786         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
30787         than an unsigned HOST_WIDE_INT.
30788         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
30789         (vect_analyze_data_ref_dependence): Change max_vf from an int *
30790         to an unsigned int *.
30791         (vect_analyze_data_ref_dependences): Likewise.
30792         (vect_compute_data_ref_alignment): Handle polynomial vf.
30793         (vect_enhance_data_refs_alignment): Likewise.
30794         (vect_prune_runtime_alias_test_list): Likewise.
30795         (vect_shift_permute_load_chain): Likewise.
30796         (vect_supportable_dr_alignment): Likewise.
30797         (dependence_distance_ge_vf): Take the vectorization factor as a
30798         poly_uint64 rather than an unsigned HOST_WIDE_INT.
30799         (vect_analyze_data_refs): Change min_vf from an int * to a
30800         poly_uint64 *.
30801         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
30802         vfm1 as a poly_uint64 rather than an int.  Make the same change
30803         for the returned bound_scalar.
30804         (vect_gen_vector_loop_niters): Handle polynomial vf.
30805         (vect_do_peeling): Likewise.  Update call to
30806         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
30807         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
30808         be constant.
30809         * tree-vect-loop.c (vect_determine_vectorization_factor)
30810         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
30811         (vect_get_known_peeling_cost): Likewise.
30812         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
30813         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
30814         (vect_transform_loop): Likewise.  Use the lowest possible VF when
30815         updating the upper bounds of the loop.
30816         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
30817         rather than an int.
30818         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
30819         polynomial unroll factors.
30820         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
30821         (vect_make_slp_decision): Likewise.
30822         (vect_supported_load_permutation_p): Likewise, and polynomial
30823         vf too.
30824         (vect_analyze_slp_cost): Handle polynomial vf.
30825         (vect_slp_analyze_node_operations): Likewise.
30826         (vect_slp_analyze_bb_1): Likewise.
30827         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
30828         than an unsigned HOST_WIDE_INT.
30829         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
30830         (vectorizable_load): Handle polynomial vf.
30831         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
30832         a poly_uint64.
30833         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
30835 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30836             Alan Hayward  <alan.hayward@arm.com>
30837             David Sherwood  <david.sherwood@arm.com>
30839         * match.pd: Handle bit operations involving three constants
30840         and try to fold one pair.
30842 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30844         * tree-vect-loop-manip.c: Include gimple-fold.h.
30845         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
30846         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
30847         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
30848         Add a path that uses a step of VF instead of 1, but disable it
30849         for now.
30850         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
30851         and niters_no_overflow parameters.  Update calls to
30852         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
30853         Create a new SSA name if the latter choses to use a ste other
30854         than zero, and return it via niters_vector_mult_vf_var.
30855         * tree-vect-loop.c (vect_transform_loop): Update calls to
30856         vect_do_peeling, vect_gen_vector_loop_niters and
30857         slpeel_make_loop_iterate_ntimes.
30858         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
30859         (vect_gen_vector_loop_niters): Update declarations after above changes.
30861 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
30863         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
30864         128-bit round to integer instructions.
30865         (ceil<mode>2): Likewise.
30866         (btrunc<mode>2): Likewise.
30867         (round<mode>2): Likewise.
30869 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30871         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
30872         unaligned VSX load/store on P8/P9.
30873         (expand_block_clear): Allow the use of unaligned VSX
30874         load/store on P8/P9.
30876 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30878         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
30879         New function.
30880         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
30881         swap associated with both a load and a store.
30883 2018-01-02  Andrew Waterman  <andrew@sifive.com>
30885         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
30886         * config/riscv/riscv.md (clear_cache): Use it.
30888 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
30890         * web.c: Remove out-of-date comment.
30892 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30894         * expr.c (fixup_args_size_notes): Check that any existing
30895         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
30896         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
30897         (emit_single_push_insn): ...here.
30899 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30901         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
30902         (const_vector_encoded_nelts): New function.
30903         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
30904         (const_vector_int_elt, const_vector_elt): Declare.
30905         * emit-rtl.c (const_vector_int_elt_1): New function.
30906         (const_vector_elt): Likewise.
30907         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
30908         of CONST_VECTOR_ELT.
30910 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30912         * expr.c: Include rtx-vector-builder.h.
30913         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
30914         directly on the tree encoding.
30915         (const_vector_from_tree): Likewise.
30916         * optabs.c: Include rtx-vector-builder.h.
30917         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
30918         sequence of "u" values.
30919         * vec-perm-indices.c: Include rtx-vector-builder.h.
30920         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
30921         directly on the vec_perm_indices encoding.
30923 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30925         * doc/rtl.texi (const_vector): Describe new encoding scheme.
30926         * Makefile.in (OBJS): Add rtx-vector-builder.o.
30927         * rtx-vector-builder.h: New file.
30928         * rtx-vector-builder.c: Likewise.
30929         * rtl.h (rtx_def::u2): Add a const_vector field.
30930         (CONST_VECTOR_NPATTERNS): New macro.
30931         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
30932         (CONST_VECTOR_DUPLICATE_P): Likewise.
30933         (CONST_VECTOR_STEPPED_P): Likewise.
30934         (CONST_VECTOR_ENCODED_ELT): Likewise.
30935         (const_vec_duplicate_p): Check for a duplicated vector encoding.
30936         (unwrap_const_vec_duplicate): Likewise.
30937         (const_vec_series_p): Check for a non-duplicated vector encoding.
30938         Say that the function only returns true for integer vectors.
30939         * emit-rtl.c: Include rtx-vector-builder.h.
30940         (gen_const_vec_duplicate_1): Delete.
30941         (gen_const_vector): Call gen_const_vec_duplicate instead of
30942         gen_const_vec_duplicate_1.
30943         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
30944         (gen_const_vec_duplicate): Use rtx_vector_builder.
30945         (gen_const_vec_series): Likewise.
30946         (gen_rtx_CONST_VECTOR): Likewise.
30947         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
30948         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
30949         Build a new vector rather than modifying a CONST_VECTOR in-place.
30950         (handle_special_swappables): Update call accordingly.
30951         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
30952         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
30953         Build a new vector rather than modifying a CONST_VECTOR in-place.
30954         (handle_special_swappables): Update call accordingly.
30956 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30958         * simplify-rtx.c (simplify_const_binary_operation): Use
30959         CONST_VECTOR_ELT instead of XVECEXP.
30961 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30963         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
30964         the selector elements to be different from the data elements
30965         if the selector is a VECTOR_CST.
30966         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
30967         ssizetype for the selector.
30969 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30971         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
30972         before testing each element individually.
30973         * tree-vect-generic.c (lower_vec_perm): Likewise.
30975 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30977         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
30978         * selftest-run-tests.c (selftest::run_tests): Call it.
30979         * vector-builder.h (vector_builder::operator ==): New function.
30980         (vector_builder::operator !=): Likewise.
30981         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
30982         (vec_perm_indices::all_from_input_p): New function.
30983         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
30984         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
30985         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
30986         instead of reading the VECTOR_CST directly.  Detect whether both
30987         vector inputs are the same before constructing the vec_perm_indices,
30988         and update the number of inputs argument accordingly.  Use the
30989         utility functions added above.  Only construct sel2 if we need to.
30991 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
30993         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
30994         the broadcast of the low byte.
30995         (expand_mult_highpart): Use an explicit encoding for the permutes.
30996         * optabs-query.c (can_mult_highpart_p): Likewise.
30997         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
30998         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30999         (vectorizable_bswap): Likewise.
31000         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
31001         explicit encoding for the power-of-2 permutes.
31002         (vect_permute_store_chain): Likewise.
31003         (vect_grouped_load_supported): Likewise.
31004         (vect_permute_load_chain): Likewise.
31006 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31008         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
31009         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
31010         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
31011         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31012         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
31013         (vect_gen_perm_mask_any): Likewise.
31015 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31017         * int-vector-builder.h: New file.
31018         * vec-perm-indices.h: Include int-vector-builder.h.
31019         (vec_perm_indices): Redefine as an int_vector_builder.
31020         (auto_vec_perm_indices): Delete.
31021         (vec_perm_builder): Redefine as a stand-alone class.
31022         (vec_perm_indices::vec_perm_indices): New function.
31023         (vec_perm_indices::clamp): Likewise.
31024         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
31025         (vec_perm_indices::new_vector): New function.
31026         (vec_perm_indices::new_expanded_vector): Update for new
31027         vec_perm_indices class.
31028         (vec_perm_indices::rotate_inputs): New function.
31029         (vec_perm_indices::all_in_range_p): Operate directly on the
31030         encoded form, without computing elided elements.
31031         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
31032         encoding.  Update for new vec_perm_indices class.
31033         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
31034         the given vec_perm_builder.
31035         (expand_vec_perm_var): Update vec_perm_builder constructor.
31036         (expand_mult_highpart): Use vec_perm_builder instead of
31037         auto_vec_perm_indices.
31038         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
31039         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
31040         or double series encoding as appropriate.
31041         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
31042         vec_perm_indices instead of auto_vec_perm_indices.
31043         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31044         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31045         (vect_permute_store_chain): Likewise.
31046         (vect_grouped_load_supported): Likewise.
31047         (vect_permute_load_chain): Likewise.
31048         (vect_shift_permute_load_chain): Likewise.
31049         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
31050         (vect_transform_slp_perm_load): Likewise.
31051         (vect_schedule_slp_instance): Likewise.
31052         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31053         (vectorizable_mask_load_store): Likewise.
31054         (vectorizable_bswap): Likewise.
31055         (vectorizable_store): Likewise.
31056         (vectorizable_load): Likewise.
31057         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
31058         vec_perm_indices instead of auto_vec_perm_indices.  Use
31059         tree_to_vec_perm_builder to read the vector from a tree.
31060         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
31061         vec_perm_builder instead of a vec_perm_indices.
31062         (have_whole_vector_shift): Use vec_perm_builder and
31063         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
31064         truncation to calc_vec_perm_mask_for_shift.
31065         (vect_create_epilog_for_reduction): Likewise.
31066         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
31067         from auto_vec_perm_indices to vec_perm_indices.
31068         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
31069         instead of changing individual elements.
31070         (aarch64_vectorize_vec_perm_const): Use new_vector to install
31071         the vector in d.perm.
31072         * config/arm/arm.c (expand_vec_perm_d::perm): Change
31073         from auto_vec_perm_indices to vec_perm_indices.
31074         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
31075         instead of changing individual elements.
31076         (arm_vectorize_vec_perm_const): Use new_vector to install
31077         the vector in d.perm.
31078         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
31079         Update vec_perm_builder constructor.
31080         (rs6000_expand_interleave): Likewise.
31081         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
31082         (rs6000_expand_interleave): Likewise.
31084 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31086         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
31087         to qimode could truncate the indices.
31088         * optabs.c (expand_vec_perm_var): Likewise.
31090 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31092         * Makefile.in (OBJS): Add vec-perm-indices.o.
31093         * vec-perm-indices.h: New file.
31094         * vec-perm-indices.c: Likewise.
31095         * target.h (vec_perm_indices): Replace with a forward class
31096         declaration.
31097         (auto_vec_perm_indices): Move to vec-perm-indices.h.
31098         * optabs.h: Include vec-perm-indices.h.
31099         (expand_vec_perm): Delete.
31100         (selector_fits_mode_p, expand_vec_perm_var): Declare.
31101         (expand_vec_perm_const): Declare.
31102         * target.def (vec_perm_const_ok): Replace with...
31103         (vec_perm_const): ...this new hook.
31104         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
31105         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
31106         * doc/tm.texi: Regenerate.
31107         * optabs.def (vec_perm_const): Delete.
31108         * doc/md.texi (vec_perm_const): Likewise.
31109         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
31110         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
31111         expand_vec_perm for constant permutation vectors.  Assert that
31112         the mode of variable permutation vectors is the integer equivalent
31113         of the mode that is being permuted.
31114         * optabs-query.h (selector_fits_mode_p): Declare.
31115         * optabs-query.c: Include vec-perm-indices.h.
31116         (selector_fits_mode_p): New function.
31117         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
31118         is defined, instead of checking whether the vec_perm_const_optab
31119         exists.  Use targetm.vectorize.vec_perm_const instead of
31120         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
31121         fit in the vector mode before using a variable permute.
31122         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
31123         vec_perm_indices instead of an rtx.
31124         (expand_vec_perm): Replace with...
31125         (expand_vec_perm_const): ...this new function.  Take the selector
31126         as a vec_perm_indices rather than an rtx.  Also take the mode of
31127         the selector.  Update call to shift_amt_for_vec_perm_mask.
31128         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
31129         Use vec_perm_indices::new_expanded_vector to expand the original
31130         selector into bytes.  Check whether the indices fit in the vector
31131         mode before using a variable permute.
31132         (expand_vec_perm_var): Make global.
31133         (expand_mult_highpart): Use expand_vec_perm_const.
31134         * fold-const.c: Includes vec-perm-indices.h.
31135         * tree-ssa-forwprop.c: Likewise.
31136         * tree-vect-data-refs.c: Likewise.
31137         * tree-vect-generic.c: Likewise.
31138         * tree-vect-loop.c: Likewise.
31139         * tree-vect-slp.c: Likewise.
31140         * tree-vect-stmts.c: Likewise.
31141         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
31142         Delete.
31143         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
31144         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
31145         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
31146         (aarch64_vectorize_vec_perm_const): ...this new function.
31147         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31148         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31149         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
31150         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
31151         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31152         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31153         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
31154         into...
31155         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
31156         check for NEON modes.
31157         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
31158         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
31159         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
31160         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
31161         into...
31162         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
31163         the old VEC_PERM_CONST conditions.
31164         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
31165         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
31166         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
31167         (ia64_vectorize_vec_perm_const_ok): Merge into...
31168         (ia64_vectorize_vec_perm_const): ...this new function.
31169         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
31170         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
31171         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
31172         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
31173         * config/mips/mips.c (mips_expand_vec_perm_const)
31174         (mips_vectorize_vec_perm_const_ok): Merge into...
31175         (mips_vectorize_vec_perm_const): ...this new function.
31176         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
31177         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
31178         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
31179         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
31180         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
31181         (rs6000_expand_vec_perm_const): Delete.
31182         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
31183         Delete.
31184         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31185         (altivec_expand_vec_perm_const_le): Take each operand individually.
31186         Operate on constant selectors rather than rtxes.
31187         (altivec_expand_vec_perm_const): Likewise.  Update call to
31188         altivec_expand_vec_perm_const_le.
31189         (rs6000_expand_vec_perm_const): Delete.
31190         (rs6000_vectorize_vec_perm_const_ok): Delete.
31191         (rs6000_vectorize_vec_perm_const): New function.
31192         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
31193         an element count and rtx array.
31194         (rs6000_expand_extract_even): Update call accordingly.
31195         (rs6000_expand_interleave): Likewise.
31196         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
31197         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
31198         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
31199         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
31200         (rs6000_expand_vec_perm_const): Delete.
31201         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
31202         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31203         (altivec_expand_vec_perm_const_le): Take each operand individually.
31204         Operate on constant selectors rather than rtxes.
31205         (altivec_expand_vec_perm_const): Likewise.  Update call to
31206         altivec_expand_vec_perm_const_le.
31207         (rs6000_expand_vec_perm_const): Delete.
31208         (rs6000_vectorize_vec_perm_const_ok): Delete.
31209         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
31210         reference to the SPE evmerge intructions.
31211         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
31212         an element count and rtx array.
31213         (rs6000_expand_extract_even): Update call accordingly.
31214         (rs6000_expand_interleave): Likewise.
31215         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
31216         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
31217         new function.
31218         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
31220 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31222         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
31223         vector mode and that that mode matches the mode of the data
31224         being permuted.
31225         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
31226         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
31227         directly using expand_vec_perm_1 when forcing selectors into
31228         registers.
31229         (expand_vec_perm_var): New function, split out from expand_vec_perm.
31231 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31233         * optabs-query.h (can_vec_perm_p): Delete.
31234         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
31235         * optabs-query.c (can_vec_perm_p): Split into...
31236         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
31237         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
31238         particular selector is valid.
31239         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31240         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31241         (vect_grouped_load_supported): Likewise.
31242         (vect_shift_permute_load_chain): Likewise.
31243         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
31244         (vect_transform_slp_perm_load): Likewise.
31245         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31246         (vectorizable_bswap): Likewise.
31247         (vect_gen_perm_mask_checked): Likewise.
31248         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
31249         implementations of variable permutation vectors into account
31250         when deciding which selector to use.
31251         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
31252         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
31253         with a false third argument.
31254         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
31255         to test whether the constant selector is valid and can_vec_perm_var_p
31256         to test whether a variable selector is valid.
31258 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31260         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
31261         * optabs-query.c (can_vec_perm_p): Likewise.
31262         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
31263         instead of vec_perm_indices.
31264         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
31265         (vect_gen_perm_mask_checked): Likewise,
31266         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
31267         (vect_gen_perm_mask_checked): Likewise,
31269 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
31271         * optabs-query.h (qimode_for_vec_perm): Declare.
31272         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
31273         (qimode_for_vec_perm): ...this new function.
31274         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
31276 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
31278         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
31279         does not have a conditional at the top.
31281 2018-01-02  Richard Biener  <rguenther@suse.de>
31283         * ipa-inline.c (big_speedup_p): Fix expression.
31285 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
31287         PR target/81616
31288         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
31289         for generic 4->6.
31291 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
31293         PR target/81616
31294         Generic tuning.
31295         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
31296         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
31297         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
31298         cond_taken_branch_cost 3->4.
31300 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
31302         PR tree-optimization/83581
31303         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
31304         TODO_cleanup_cfg if any changes have been made.
31306         PR middle-end/83608
31307         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
31308         convert_modes if target mode has the right side, but different mode
31309         class.
31311         PR middle-end/83609
31312         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
31313         last argument when extracting from CONCAT.  If either from_real or
31314         from_imag is NULL, use expansion through memory.  If result is not
31315         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
31316         the parts directly to inner mode, if even that fails, use expansion
31317         through memory.
31319         PR middle-end/83623
31320         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
31321         check for bswap in mode rather than HImode and use that in expand_unop
31322         too.
31324 Copyright (C) 2018 Free Software Foundation, Inc.
31326 Copying and distribution of this file, with or without modification,
31327 are permitted in any medium without royalty provided the copyright
31328 notice and this notice are preserved.