ux.texi: move "Quoting" and "Fix-it hints" from DiagnosticsGuidelines wiki page
[official-gcc.git] / gcc / ChangeLog
blobb649da16184bf5d9349f39eef2cf59cb85762bce
1 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2             Martin Sebor  <msebor@redhat.com>
3             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
5         * doc/ux.texi (Quoting): New subsection, adapted from material at
6         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
7         MartinSebor and ManuelLopezIbanez.
8         (Fix-it hints): Note that fix-it hints shouldn't be marked for
9         translation.
11 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
13         PR middle-end/87469
14         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
15         max value.
17 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
19         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
21 2018-10-29  Paul Koning  <ni1d@arrl.net>
23         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
25 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
27         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
28         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
29         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
30         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
31         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
32         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
33         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
34         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
35         Change 'vector' to '__vector'.
36         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
37         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
38         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
39         _mm_avg_pu16): Likewise.  And, whitespace corrections.
41 2018-10-29  Richard Biener  <rguenther@suse.de>
43         PR tree-optimization/87785
44         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
45         internal defs.
47 2018-10-29  Olivier Hainque  <hainque@adacore.com>
49         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
51 2018-10-29  Olivier Hainque  <hainque@adacore.com>
53         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
54         ports configuration macro, defaults to "gnu".
55         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
56         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
58 2018-10-29  Olivier Hainque  <hainque@adacore.com>
60         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
62 2018-10-29  Richard Biener  <rguenther@suse.de>
64         PR tree-optimization/87790
65         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
66         (vect_make_slp_decision): Adjust.
67         (vect_slp_analyze_bb_1): Likewise.
68         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
69         edges.
71 2018-10-29  Richard Biener  <rguenther@suse.de>
73         PR tree-optimization/87785
74         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
75         and processing.
76         (vect_build_slp_tree): Likewise.
77         (vect_gather_slp_loads): New function.
78         (vect_analyze_slp_instance): Gather loads separately from the
79         SLP tree build.
81 2018-10-29  Martin Liska  <mliska@suse.cz>
83         * Makefile.in: Make dependency to json.o.
84         * doc/gcov.texi: Document new JSON format, remove
85         old intermediate format documentation.
86         * gcov.c (struct function_info): Come up with m_name and
87         m_demangled_name.
88         (function_info::function_info): Initialize it.
89         (function_info::~function_info): Release it.
90         (main): Rename flag_intermediate_format to flag_json_format.
91         (print_usage): Describe --json-format.
92         (process_args): Set flag_json_format.
93         (output_intermediate_line): Remove.
94         (output_intermediate_json_line): Likewise.
95         (get_gcov_intermediate_filename): Return new extension
96         ".gcov.json.gz".
97         (output_intermediate_file): Implement JSON emission.
98         (output_json_intermediate_file): Implement JSON emission.
99         (generate_results): Use ::get_name for function name.
100         Handle JSON output file.
101         (read_graph_file): Use ::get_name instead of cplus_demangle.
102         (read_count_file): Likewise.
103         (solve_flow_graph): Likewise.
104         (add_line_counts): Likewise.
105         (accumulate_line_counts): Use new flag_json_format.
106         (output_function_details): Use ::get_name instead of cplus_demangle.
107         (output_lines): Likewise.
108         * json.cc (test_writing_literals): Add new tests.
109         * json.h (class literal): Add new boolean constructor.
111 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
113         PR rtl-optimization/87701
114         PR rtl-optimization/87780
115         * combine.c (make_more_copies): Rewrite.
117 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
119         * doc/generic.texi (ABSU_EXPR): Document.
120         * match.pd (absu(x)*absu(x) -> x*x): Handle.
121         (absu(absu(X)) -> absu(X)): Likewise.
122         (absu(-X) -> absu(X)): Likewise.
123         (absu(X)  where X is nonnegative -> X): Likewise.
125 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
127         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
128         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
129         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
130         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
131         (s-tm-texi): Also check timestamp on d-target.def.
132         (generated_files): Add TM_D_H and d-target-hooks-def.h.
133         (build/genhooks.o): Also depend on D_TARGET_DEF.
134         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
135         variables.
136         * config/aarch64/aarch64-d.c: New file.
137         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
138         Define.
139         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
140         prototype.
141         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
142         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
143         * config/arm/arm-d.c: New file.
144         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
145         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
146         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
147         * config/arm/t-arm (arm-d.o): New rule.
148         * config/default-d.c: New file.
149         * config/glibc-d.c: New file.
150         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
151         * config/i386/i386-d.c: New file.
152         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
153         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
154         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
155         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
156         * config/i386/t-i386 (i386-d.o): New rule.
157         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
158         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
159         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
160         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
161         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
162         * config/mips/mips-d.c: New file.
163         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
164         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
165         * config/mips/t-mips (mips-d.o): New rule.
166         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
167         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
168         * config/powerpcspe/powerpcspe-d.c: New file.
169         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
170         New prototype.
171         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
172         Support GNU D by using 0 as the language type.
173         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
174         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
175         * config/riscv/riscv-d.c: New file.
176         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
177         prototype.
178         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
179         * config/riscv/t-riscv (riscv-d.o): New rule.
180         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
181         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
182         * config/rs6000/rs6000-d.c: New file.
183         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
184         prototype.
185         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
186         Support GNU D by using 0 as the language type.
187         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
188         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
189         * config/s390/s390-d.c: New file.
190         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
191         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
192         * config/s390/t-s390 (s390-d.o): New rule.
193         * config/sparc/sparc-d.c: New file.
194         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
195         prototype.
196         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
197         * config/sparc/t-sparc (sparc-d.o): New rule.
198         * config/t-glibc (glibc-d.o): New rule.
199         * configure: Regenerated.
200         * configure.ac (tm_d_file): New variable.
201         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
202         * doc/contrib.texi (Contributors): Add self for the D frontend.
203         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
204         * doc/install.texi (Configuration): Mention libphobos as an option for
205         --enable-shared.  Mention d as an option for --enable-languages.
206         (Testing): Mention check-d as a target.
207         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
208         name suffixes.  Mention d as a -x option.
209         * doc/sourcebuild.texi (Top Level): Mention libphobos.
210         * doc/standards.texi (Standards): Add section on D language.
211         * doc/tm.texi: Regenerated.
212         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
213         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
214         * dwarf2out.c (is_dlang): New function.
215         (gen_compile_unit_die): Use DW_LANG_D for D.
216         (declare_in_namespace): Return module die for D, instead of adding
217         extra declarations into the namespace.
218         (gen_namespace_die): Generate DW_TAG_module for D.
219         (gen_decl_die): Handle CONST_DECLSs for D.
220         (dwarf2out_decl): Likewise.
221         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
222         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
223         kinds of aggregates.
224         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
225         * genhooks.c: Include d/d-target.def.
227 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
229         PR target/85669
230         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
231         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
232         (STACK_DYNAMIC_OFFSET): Likewise.
234 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
236         PR target/80024
237         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
238         error message.
240 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
242         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
243         return type and other typos.
245 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
247         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
248         a shorter sequence with fewer branches.
249         (emit_final_str_compare_gpr): Ditto.
251 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
253         * config/rs6000/tmmintrin.h: New file.
254         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
256 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
258         * config/rs6000/mmintrin.h: Enable 32bit compilation.
259         * config/rs6000/xmmintrin.h: Likewise.
261 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
263         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
265 2018-10-26  Richard Biener  <rguenther@suse.de>
267         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
268         and wrapper.
269         (vect_mark_slp_stmts_relevant): Likewise.
270         (vect_detect_hybrid_slp_stmts): Likewise.
271         (vect_bb_slp_scalar_cost): Likewise.
272         (vect_remove_slp_scalar_calls): Likewise.
274 2018-10-26  Jan Hubicka  <jh@suse.cz>
276         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
277         (warn_types_mismatch): Fix walk of DECL_NAME.
278         (odr_types_equivalent_p): Fix overactive assert.
280 2018-10-26  Richard Biener  <rguenther@suse.de>
282         PR tree-optimization/87105
283         * tree-vectorizer.h (_slp_tree::refcnt): New member.
284         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
285         refcnt.
286         (vect_create_new_slp_node): Initialize refcnt to one.
287         (bst_traits): Move.
288         (scalar_stmts_set_t, bst_fail): Remove.
289         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
290         (vect_build_slp_tree): Add bst_map argument and lookup
291         already created SLP nodes.
292         (vect_print_slp_tree): Handle a SLP graph, print SLP node
293         addresses.
294         (vect_slp_rearrange_stmts): Handle a SLP graph.
295         (vect_analyze_slp_instance): Adjust and free SLP nodes from
296         the CSE map.  Fix indenting.
297         (vect_schedule_slp_instance): Add short-cut.
299 2018-10-26  Martin Liska  <mliska@suse.cz>
301         PR testsuite/86158
302         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
303         addr_expr and not with pointers.
305 2018-10-26  Jan Hubicka  <jh@suse.cz>
307         * tree.c (free_lang_data_in_type): Only check main variants.
308         * ipa-devirt.c (warn_odr): Make static.
309         (types_same_for_odr): Drop strict variant.
310         (types_odr_comparable): Likewise.
311         (odr_or_derived_type_p): Look for main variants.
312         (odr_name_hasher::equal): Cleanup comment.
313         (odr_subtypes_equivalent): Add warn and warned arguments; check main
314         variants.
315         (type_variants_equivalent_p): break out from ...
316         (odr_types_equivalent): ... here; go for main variants where needed.
317         (warn_odr): ... here; turn static.
318         (warn_types_mismatch): Compare mangled names of main variants.
319         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
320         (type_with_linkage_p): Sanity check that we look at main variant.
321         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
322         * tree.h (types_same_for_odr): Drop strict argument.
324 2018-10-26  Richard Biener  <rguenther@suse.de>
326         PR tree-optimization/87746
327         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
328         Simplify and fix WRT strided store groups with size not
329         equal to step in element count.
330         (vect_analyze_group_access_1): Dump the whole group.
332 2018-10-25  Carl Love  <cel@us.ibm.com>
334         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
335         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
336         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
337         precicion entry for each overloaded builtin.
338         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
339         VSCEDPUO): Rename overloaded name.
340         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
341         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
342         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
343         define_expand for xscmpexqp instruction.
344         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
346 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
347             Jinsong Ji <jji@us.ibm.com>
349         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
350         function with vec_sl.
351         (_mm_slli_epi32): Likewise.
352         (_mm_slli_epi64): Likewise.
353         (_mm_srai_epi16): Replace deprecated function with vec_sra.
354         (_mm_srai_epi32): Likewise.
355         (_mm_srli_epi16): Replace deprecated function with vec_sr.
356         (_mm_srli_epi32): Likewise.
357         (_mm_srli_epi64): Likewise.
358         (_mm_sll_epi16): Replace deprecated function with vec_sl.
359         (_mm_sll_epi32): Likewise.
360         (_mm_sll_epi64): Likewise.
361         (_mm_sra_epi16): Replace deprecated function with vec_sra.
362         (_mm_sra_epi32): Likewise.
363         (_mm_srl_epi16): Replace deprecated function with vec_sr.
364         (_mm_srl_epi32): Likewise.
365         (_mm_srl_epi64): Likewise.
367 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
368             Jinsong Ji <jji@us.ibm.com>
370         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
371         comparison operators with vec_cmp* for compatibility due to
372         unfortunate history; clean up formatting and use types more
373         appropriately.
374         (_mm_sll_epi32): Likewise.
375         (_mm_sll_epi64): Likewise.
376         (_mm_srl_epi16): Likewise.
377         (_mm_srl_epi32): Likewise.
378         (_mm_srl_epi64): Likewise.
380 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
381             Jinsong Ji <jji@us.ibm.com>
383         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
384         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
385         __vector __bool int.  Use vec_cmpgt in preference to deprecated
386         function vec_vcmpgtfp.
387         (_mm_max_ps): Likewise.
389 2018-10-25  Jeff Law  <law@redhat.com>
391         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
392         if returning false.
394 2018-10-25  Martin Sebor  <msebor@redhat.com>
396         * doc/extend.texi (aligned): Expand attribute description.
397         (Alignment): Rename section.  Discuss function arguments.
399 2018-10-25  Jan Hubicka  <jh@suse.cz>
401         * ipa-devirt.c (main_odr_variant): Remove.
402         (hash_odr_name, types_same_for_odr, types_odr_comparable,
403         odr_name_hasher::equal, odr_subtypes_equivalent_p):
404         Drop use of main_odr_variant.
405         (add_type_duplicate): Silence confused warnings on integer types.
406         (get_odr_type): Always look for main variant.
407         (register_odr_type): Simplify.
409 2018-10-25  Richard Biener  <rguenther@suse.de>
411         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
412         Initialize ng to silence error with release checking bootstrap.
414 2018-10-25  Richard Biener  <rguenther@suse.de>
416         * tree-if-conv.c: Include tree-ssa-sccvn.h.
417         (tree_if_conversion): Run CSE on the if-converted loop body.
419 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
421         * config/s390/constraints.md (ZL): New constraint.
422         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
423         operands.
424         * config/s390/s390.md (movdi_larl): Remove.
425         (movdi_64): Add the LARL alternative.
427 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
429         PR bootstrap/87747
430         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
431         (RTX_CODE_HWINT_P): New macro.
432         (rtx_code_size): Use RTX_CODE_HWINT_P ().
434 2018-10-25  Jan Hubicka  <jh@suse.cz>
436         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
437         is anonymous.
439 2018-10-25  Richard Biener  <rguenther@suse.de>
441         PR tree-optimization/87665
442         PR tree-optimization/87745
443         * tree-vectorizer.h (get_earlier_stmt): Remove.
444         (get_later_stmt): Pick up UID from the original non-pattern stmt.
446 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
448         * options.texi (Deprecated): Move list to Var section.
450 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
451             Jinsong Ji <jji@us.ibm.com>
453         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
454         __vector long to __vector long long.
455         (_mm_cvtpd_ps): Likewise.
456         (_mm_cvttpd_epi32): Likewise.
457         (_mm_cvtpi32_pd): Likewise.
458         (_mm_unpackhi_epi64): Likewise.
459         (_mm_unpacklo_epi64): Likewise.
461 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
463         PR rtl-optimization/87720
464         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
466 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
468         * gimple-ssa-isolate-paths.c
469         (find_implicit_erroneous_behavior): Do not change code if the
470         pass is running for warnings only.
471         (find_explicit_erroneous_behavior): Likewise.
473 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
475         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
476         Define as rs6000_mangle_decl_assembler_name.
477         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
478         long double to IEEE long double, switch the names of the long
479         double built-in functions to be <func>f128 instead of <func>l.
481 2018-10-24  Martin Sebor  <msebor@redhat.com>
483         * doc/extend.texi (nonnull): List no-argument form.  Reference
484         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
486 2018-10-24  Richard Biener  <rguenther@suse.de>
488         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
490 2018-10-24  Martin Liska  <mliska@suse.cz>
492         PR tree-optimization/84436
493         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
494         Remove.
495         (switch_conversion::contains_linear_function_p): New.
496         (switch_conversion::build_one_array): Support linear
497         transformation on input.
498         * tree-switch-conversion.h (struct switch_conversion): Add
499         contains_linear_function_p declaration.
501 2018-10-24  Richard Biener  <rguenther@suse.de>
503         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
504         if its argument is CONSTANT_CLASS_P.
506 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
508         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
509         it is wrong for forward declarations.
511 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
513         * config/s390/s390.c (s390_check_qrst_address): Add the missing
514         SYMBOL_REF_P () check.
516 2018-10-24  Richard Biener  <rguenther@suse.de>
518         PR tree-optimization/87105
519         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
520         dump classification.
521         (vect_analyze_data_ref_accesses): Handle duplicate loads and
522         stores by splitting the affected group after the fact.
523         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
524         fail the SLP build because of size constraints.
526 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
528         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
529         * configure: Regenerate.
530         * config.in: Regenerate.
531         * varasm.c (mergeable_string_section): Use readonly_data_section
532         if linker doesn't support SHF_MERGE with alignment > 8.
533         (mergeable_constant_section): Likewise.
535 2018-10-24  Richard Biener  <rguenther@suse.de>
537         PR tree-optimization/84013
538         * tree-ssa-structalias.c (struct msdi_data): New struct for
539         marshalling data to walk_stmt_load_store_ops.
540         (maybe_set_dependence_info): Refactor as callback for
541         walk_stmt_load_store_ops.
542         (compute_dependence_clique): Set restrict info on all stmt kinds.
544 2018-10-24  Martin Liska  <mliska@suse.cz>
546         * cgraph.c (cgraph_node::dump):
547         Remove reduntant dumps and make tp_first_run dump more compact.
549 2018-10-24  Richard Biener  <rguenther@suse.de>
551         PR tree-optimization/87665
552         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
553         to reflect reality.
555 2018-10-12  Jeff Law  <law@redhat.com>
557         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
558         for H8/S.
560 2018-10-23  Richard Biener  <rguenther@suse.de>
562         * tree-vrp.c (add_assert_info): Guard dump_printf with
563         dump_enabled_p.
564         * gimple-ssa-evrp-analyze.c
565         (evrp_range_analyzer::record_ranges_from_incoming_edge):
566         Use value_range::ignore_equivs_equal_p.
568 2018-10-23  Richard Biener  <rguenther@suse.de>
570         PR tree-optimization/87105
571         PR tree-optimization/87608
572         * passes.def (pass_all_early_optimizations): Add early phi-opt
573         after dce.
574         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
575         addition to debug stmts.
576         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
577         and abs replacement early.
578         * tree-cfg.c (gimple_empty_block_p): Likewise.
580 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
582         PR target/86383
583         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
584         specified to configure.
585         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
587 2018-10-23  Richard Biener  <rguenther@suse.de>
589         PR tree-optimization/87700
590         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
592 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
594         PR target/87674
595         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
596         second argument from __mmask16 to __mmask8.
597         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
598         _mm_mask_packs_epi32): Likewise.
599         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
600         Likewise.
601         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
603 2018-10-23  Richard Biener  <rguenther@suse.de>
605         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
607 2018-10-23  Richard Biener  <rguenther@suse.de>
609         PR tree-optimization/86144
610         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
611         over simd attribute.
613 2018-10-23  Richard Biener  <rguenther@suse.de>
615         PR tree-optimization/87693
616         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
617         the case we do not find the taken edge.
619 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
620             Jinsong Ji  <jji@us.ibm.com>
622         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
623         (_mm_store_pd): Use unaligned vector type for pointer cast.
624         (_mm_maskmoveu_si128): Likewise.
625         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
626         (_mm_store_ps): Use unaligned vector type for pointer cast.
628 2018-10-22  Paul Koning  <ni1d@arrl.net>
630         * symtab.c (symtab_node::increase_alignment): Correct max
631         alignment check.
633 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
635         PR tree-optimization/87633
636         * match.pd: Do not generate unordered integer comparisons.
638 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
640         PR rtl-optimization/87600
641         * combine.c: Add include of expr.h.
642         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
643         register to a pseudo.
644         (make_more_copies): New function, add a copy to a new pseudo after
645         the moves from hard registers into pseudos.
646         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
647         later.  Call make_more_copies.
649 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
651         * lra-constraints.c (process_alt_operands): New local array,
652         matching_early_clobber.  Check matching_early_clobber before
653         decrementing reject, and set matching_early_clobber after.
655 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
657         PR target/87598
658         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
659         call output_addr_const and hope for the best.
661 2018-10-22  Richard Biener  <rguenther@suse.de>
663         * gimple-ssa-evrp-analyze.c
664         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
665         smarter about what ranges to use.
666         * tree-vrp.c (add_assert_info): Dump here.
667         (register_edge_assert_for_2): Instead of here at multiple but
668         not all places.
670         * gcc.dg/tree-ssa/evrp12.c: New testcase.
671         * gcc.dg/predict-6.c: Adjust.
672         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
673         * gcc.dg/tree-ssa/vrp02.c: Likewise.
674         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
676 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
677         Richard Biener  <rguenther@suse.de>
679         * bitmap.h: Update data structure documentation, including a
680         description of bitmap views as either linked-lists or splay trees.
681         (struct bitmap_element_def): Update comments for splay tree bitmaps.
682         (struct bitmap_head_def): Likewise.
683         (bitmap_list_view, bitmap_tree_view): New prototypes.
684         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
685         tree_form fields.
686         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
687         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
688         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
689         released bitmap element here.
690         (bitmap_element_free): Remove.
691         (bitmap_elt_clear_from): Work on splay tree bitmaps.
692         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
693         this function similar ones such that linked-list bitmap implementation
694         functions are grouped.
695         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
696         and moved for grouping.
697         (bitmap_list_insert_element_after): Renamed from
698         bitmap_elt_insert_after, and moved for grouping.
699         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
700         (bitmap_tree_link_left, bitmap_tree_link_right,
701         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
702         bitmap_tree_link_element, bitmap_tree_unlink_element,
703         bitmap_tree_find_element): New functions for splay-tree bitmap
704         implementation.
705         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
706         Renamed and moved, see above entries.
707         (bitmap_tree_listify_from): New function to convert part of a splay
708         tree bitmap to a linked-list bitmap.
709         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
710         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
711         (bitmap_find_bit): Remove.
712         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
713         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
714         Handle splay tree bitmaps.
715         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
716         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
717         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
718         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
719         bitmap_intersect_compl_p, bitmap_ior_and_compl,
720         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
721         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
722         corresponding changes to use linked-list specific bitmap_element
723         manipulation functions as applicable for efficiency.
724         (bitmap_tree_to_vec): New function.
725         (debug_bitmap_elt_file): New function split out from ...
726         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
727         (bitmap_print): Likewise.
729         PR tree-optimization/63155
730         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
731         SSA edge worklists.
732         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
733         in tree-view.
735 2018-10-22  Martin Liska  <mliska@suse.cz>
737         PR tree-optimization/87686
738         Revert
739         2018-08-29  Martin Liska  <mliska@suse.cz>
741         * tree-switch-conversion.c (switch_conversion::expand):
742         Strenghten assumption about gswitch statements.
744 2018-10-22  Martin Liska  <mliska@suse.cz>
746         * ipa-icf.c (sem_item::compare_attributes): Remove.
747         (sem_item::compare_referenced_symbol_properties): Use
748         attribute_list_equal instead.
749         (sem_function::equals_wpa): Likewise.
750         * ipa-icf.h: Remove compare_attributes.
752 2018-10-22  Richard Biener  <rguenther@suse.de>
754         PR middle-end/87682
755         * mem-stats.h (mem_usage::operator==): Fix pasto.
757 2018-10-22  Richard Biener  <rguenther@suse.de>
759         PR tree-optimization/87640
760         * tree-vrp.c (set_value_range_with_overflow): Decompose
761         incomplete result.
762         (extract_range_from_binary_expr_1): Adjust.
764 2018-10-22  Martin Jambor  <mjambor@suse.cz>
766         * tree-eh.h (stmt_could_throw_p): Add function parameter.
767         (stmt_can_throw_external): Likewise.
768         (stmt_can_throw_internal): Likewise.
769         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
770         (lower_eh_constructs_2): Likewise.
771         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
772         (stmt_can_throw_external): Likewise.
773         (stmt_can_throw_internal): Likewise.
774         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
775         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
776         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
777         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
778         (pass_lower_eh_dispatch::execute): Pass cfun to
779         stmt_can_throw_external.
780         (cleanup_empty_eh): Likewise.
781         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
782         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
783         stmt_can_throw_external instead of pushing it to cfun.
784         (symbol_table::create_edge): Likewise.
785         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
786         stmt_can_throw_internal.
787         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
788         to stmt_could_throw_p.
789         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
790         stmt_can_throw_internal.
791         (pass_store_merging::execute): Likewise.
792         * gimple-ssa-strength-reduction.c
793         (find_candidates_dom_walker::before_dom_children): Pass cfun to
794         stmt_could_throw_p.
795         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
796         stmt_can_throw_internal.
797         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
798         to stmt_can_throw_external.
799         (check_stmt): Pass cfun to stmt_could_throw_p.
800         (check_stmt): Pass cfun to stmt_can_throw_external.
801         (pass_nothrow::execute): Likewise.
802         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
803         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
804         stmt_can_throw_internal.
805         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
806         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
807         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
808         * tree-complex.c (expand_complex_libcall): Pass cfun to
809         stmt_could_throw_p and to stmt_can_throw_internal.
810         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
811         * tree-inline.c (copy_edges_for_bb): Likewise.
812         (maybe_move_debug_stmts_to_successors): Likewise.
813         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
814         stmt_could_throw_p.
815         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
816         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
817         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
818         stmt_can_throw_internal.
819         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
820         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
821         stmt_could_throw_p.
822         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
823         stmt_can_throw_internal.
824         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
825         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
826         stmt_could_throw_p.
827         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
828         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
829         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
830         (convert_mult_to_fma_1): Likewise.
831         (convert_to_divmod): Likewise.
832         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
833         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
834         * tree-ssa-propagate.c
835         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
836         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
837         (maybe_optimize_range_tests): Likewise.
838         (linearize_expr_tree): Likewise.
839         (reassociate_bb): Likewise.
840         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
841         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
842         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
843         (handle_char_store): Likewise.
844         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
845         stmt_can_throw_internal.
846         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
847         stmt_could_throw_p.
848         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
849         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
850         (vectorizable_simd_clone_call): Likewise.
851         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
852         (gimple_stringop_fixed_value): Likewise.
854 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
856         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
857         literal pool references.
858         (s390_check_qrst_address): Adapt to the new behavior of
859         s390_loadrelative_operand_p ().
861 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
863         PR target/72782
864         * config/i386/sse.md (*andnot<mode>3_bcst): New.
866 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
868         PR target/72782
869         * config/i386/sse.md (*<code><mode>3_bcst): New.
871 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
873         PR target/72782
874         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
875         V4DI, V16SI and V8DI.
876         (*sub<mode>3<mask_name>_bcst): New.
877         (*add<mode>3<mask_name>_bcst): Likewise.
879 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
880             Jinsong Ji  <jji@us.ibm.com>
882         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
883         __m64 with __vector unsigned long long for compatibility.
884         (_mm_movemask_epi8): Likewise.
885         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
886         (_mm_cvttps_pi32): Likewise.
887         (_mm_cvtpi32_ps): Likewise.
888         (_mm_cvtps_pi16): Likewise.
889         (_mm_loadh_pi): Likewise.
890         (_mm_storeh_pi): Likewise.
891         (_mm_movehl_ps): Likewise.
892         (_mm_movelh_ps): Likewise.
893         (_mm_loadl_pi): Likewise.
894         (_mm_storel_pi): Likewise.
895         (_mm_movemask_ps): Likewise.
896         (_mm_shuffle_pi16): Likewise.
898 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
900         PR target/72782
901         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
902         __builtin_ia32_vfnmsubpd512_mask.
903         (_mm512_mask_fnmsub_round_pd): Likewise.
904         (_mm512_fnmsub_pd): Likewise.
905         (_mm512_mask_fnmsub_pd): Likewise.
906         (_mm512_maskz_fnmsub_round_pd): Use
907         __builtin_ia32_vfnmsubpd512_maskz.
908         (_mm512_maskz_fnmsub_pd): Likewise.
909         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
910         (_mm512_mask_fnmsub_round_ps): Likewise.
911         (_mm512_fnmsub_ps): Likewise.
912         (_mm512_mask_fnmsub_ps): Likewise.
913         (_mm512_maskz_fnmsub_round_ps): Use
914         __builtin_ia32_vfnmsubps512_maskz.
915         (_mm512_maskz_fnmsub_ps): Likewise.
916         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
917         __builtin_ia32_vfnmsubpd256_mask.
918         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
919         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
920         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
921         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
922         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
923         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
924         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
925         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
926         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
927         __builtin_ia32_vfnmsubpd.
928         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
929         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
930         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
931         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
932         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
933         * config/i386/i386-builtin.def: Add
934         __builtin_ia32_vfnmsubpd256_mask,
935         __builtin_ia32_vfnmsubpd256_maskz,
936         __builtin_ia32_vfnmsubpd128_mask,
937         __builtin_ia32_vfnmsubpd128_maskz,
938         __builtin_ia32_vfnmsubps256_mask,
939         __builtin_ia32_vfnmsubps256_maskz,
940         __builtin_ia32_vfnmsubps128_mask,
941         __builtin_ia32_vfnmsubps128_maskz,
942         __builtin_ia32_vfnmsubpd512_mask,
943         __builtin_ia32_vfnmsubpd512_maskz,
944         __builtin_ia32_vfnmsubps512_mask,
945         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
946         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
947         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
948         __builtin_ia32_vfnmsubpd256.
949         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
950         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
951         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
952         Likewise.
953         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
954         Likewise.
955         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
956         Likewise.
957         (fmai_vmfnmsub_<mode><round_name>): Likewise.
959 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
961         PR target/72782
962         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
963         __builtin_ia32_vfnmaddpd512_mask.
964         (_mm512_mask_fnmadd_round_pd): Likewise.
965         (_mm512_fnmadd_pd): Likewise.
966         (_mm512_mask_fnmadd_pd): Likewise.
967         (_mm512_maskz_fnmadd_round_pd): Use
968         __builtin_ia32_vfnmaddpd512_maskz.
969         (_mm512_maskz_fnmadd_pd): Likewise.
970         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
971         (_mm512_mask_fnmadd_round_ps): Likewise.
972         (_mm512_fnmadd_ps): Likewise.
973         (_mm512_mask_fnmadd_ps): Likewise.
974         (_mm512_maskz_fnmadd_round_ps): Use
975         __builtin_ia32_vfnmaddps512_maskz.
976         (_mm512_maskz_fnmadd_ps): Likewise.
977         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
978         __builtin_ia32_vfnmaddpd256_mask.
979         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
980         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
981         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
982         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
983         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
984         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
985         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
986         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
987         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
988         __builtin_ia32_vfnmaddpd.
989         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
990         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
991         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
992         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
993         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
994         * config/i386/i386-builtin.def: Add
995         __builtin_ia32_vfnmaddpd256_mask,
996         __builtin_ia32_vfnmaddpd256_maskz,
997         __builtin_ia32_vfnmaddpd128_mask,
998         __builtin_ia32_vfnmaddpd128_maskz,
999         __builtin_ia32_vfnmaddps256_mask,
1000         __builtin_ia32_vfnmaddps256_maskz,
1001         __builtin_ia32_vfnmaddps128_mask,
1002         __builtin_ia32_vfnmaddps128_maskz,
1003         __builtin_ia32_vfnmaddpd512_mask,
1004         __builtin_ia32_vfnmaddpd512_maskz,
1005         __builtin_ia32_vfnmaddps512_mask,
1006         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
1007         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
1008         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
1009         __builtin_ia32_vfnmaddpd256.
1010         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
1011         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
1012         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
1013         Likewise.
1014         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
1015         Likewise.
1016         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
1017         Likewise.
1018         (fmai_vmfnmadd_<mode><round_name>): Likewise.
1020 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1022         PR target/72782
1023         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
1024         __builtin_ia32_vfmsubpd512_mask.
1025         (_mm512_mask_fmsub_round_pd): Likewise.
1026         (_mm512_fmsub_pd): Likewise.
1027         (_mm512_mask_fmsub_pd): Likewise.
1028         (_mm512_maskz_fmsub_round_pd): Use
1029         __builtin_ia32_vfmsubpd512_maskz.
1030         (_mm512_maskz_fmsub_pd): Likewise.
1031         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
1032         (_mm512_mask_fmsub_round_ps): Likewise.
1033         (_mm512_fmsub_ps): Likewise.
1034         (_mm512_mask_fmsub_ps): Likewise.
1035         (_mm512_maskz_fmsub_round_ps): Use
1036         __builtin_ia32_vfmsubps512_maskz.
1037         (_mm512_maskz_fmsub_ps): Likewise.
1038         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
1039         __builtin_ia32_vfmsubpd256_mask.
1040         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
1041         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
1042         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
1043         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1044         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
1045         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
1046         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
1047         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
1048         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
1049         __builtin_ia32_vfmsubpd.
1050         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
1051         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
1052         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
1053         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
1054         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
1055         * config/i386/i386-builtin.def: Add
1056         __builtin_ia32_vfmsubpd256_mask,
1057         __builtin_ia32_vfmsubpd256_maskz,
1058         __builtin_ia32_vfmsubpd128_mask,
1059         __builtin_ia32_vfmsubpd128_maskz,
1060         __builtin_ia32_vfmsubps256_mask,
1061         __builtin_ia32_vfmsubps256_maskz,
1062         __builtin_ia32_vfmsubps128_mask,
1063         __builtin_ia32_vfmsubps128_maskz,
1064         __builtin_ia32_vfmsubpd512_mask,
1065         __builtin_ia32_vfmsubpd512_maskz,
1066         __builtin_ia32_vfmsubps512_mask,
1067         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
1068         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
1069         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
1070         __builtin_ia32_vfmsubpd256.
1071         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
1072         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
1073         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
1074         Likewise.
1075         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
1076         Likewise.
1077         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
1078         Likewise.
1079         (fmai_vmfmsub_<mode><round_name>): Likewise.
1081 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1083         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
1084         Remove plus.  Renamed to ...
1085         (*sub<mode>3<mask_name>_bcst): This.
1086         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
1087         (*add<mode>3<mask_name>_bcst): This.
1089 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1091         PR target/72782
1092         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
1094 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
1096         PR target/87662
1097         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
1098         (_mm_or_epi32): Likewise.
1099         (_mm256_xor_epi32): Likewise.
1100         (_mm_xor_epi32): Likewise.
1101         (_mm256_or_epi64): Likewise.
1102         (_mm_or_epi64): Likewise.
1103         (_mm256_xor_epi64): Likewise.
1104         (_mm_xor_epi64): Likewise.
1106 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
1108         PR target/72782
1109         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
1111 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
1113         PR middle-end/87647
1114         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
1116 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
1118         * doc/ux.texi: Move @section directly after @node.
1120 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
1122         PR middle-end/85488
1123         PR middle-end/87649
1124         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
1125         depend closely nested inside of loop with ordered clause with
1126         a parameter.
1128 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
1130         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
1131         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
1132         * doc/ux.texi: New file.
1134 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
1136         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
1137         be the first CR field allocated.
1139 2018-10-19  Richard Biener  <rguenther@suse.de>
1141         PR target/87657
1142         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
1143         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
1145 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
1147         PR target/72782
1148         * config/i386/sse.md
1149         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
1150         (*add<mode>3<mask_name>_bcst_2): Likewise.
1152 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
1154         * config/i386/sse.md
1155         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1156         Replace nonimmediate_operand with register_operand.
1157         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1158         Likewise.
1159         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1160         Likewise.
1162 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
1164         PR rtl-optimization/87596
1165         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
1166         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
1167         for instructions in FROM..TO range.
1169 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
1171         * cfgexpand.c (expand_one_var): Use specific wording in error message
1172         for non-local frame variables.
1173         * stor-layout.c (layout_decl): Do not issue a warning for them.
1175 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
1177         * haifa-sched.c (priority): Add force_recompute parameter.
1178         (apply_replacement): Call priority () with force_recompute = true.
1179         (restore_pattern): Likewise.
1181 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1183         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
1184         HOST_BITS_PER_WIDE_INT.
1185         (test_vector_ops_duplicate): Likewise.
1187 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1189         PR target/72782
1190         * config/i386/sse.md (VF_AVX512): New.
1191         (avx512bcst): Likewise.
1192         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
1193         Likewise.
1194         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
1195         Likewise.
1196         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
1197         Likewise.
1199 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
1201         * doc/invoke.texi (-dumpversion): Improve grammar.
1202         (-dumpfullversion): Make more consistent with -dumpversion.
1204 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
1206         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
1207         Set JUMP_LABEL to the jump insn.
1208         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
1209         Predict emitted jump and add label to jump insn.
1211 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
1213         PR tree-optimization/87562
1214         * input.c (get_substring_ranges_for_loc): Use
1215         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
1216         getting the linemap for the endpoint.  Verify that it's either
1217         in the same linemap as the start point's spelling location, or
1218         at least in the same file.
1220 2018-10-18  Richard Biener  <rguenther@suse.de>
1222         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
1223         feed width-specific load/store costs through ix86_vec_cost.
1224         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
1225         (k8_cost): Likewise.
1226         (bdver_cost): Likewise.
1227         (znver1_cost): Likewise.
1228         (btver1_cost): Likewise.
1229         (btver2_cost): Likewise.
1231 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1233         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
1234         to simplify subreg of vec_merge.
1236 2018-10-18  Richard Biener  <rguenther@suse.de>
1238         * config/i386/i386.c: Fix costing of vector FMA.
1240 2018-10-18  Richard Biener  <rguenther@suse.de>
1242         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
1243         and argument.
1244         (ix86_builtin_vectorization_cost): For vec_construct properly
1245         cost insertion into SSE regs.
1246         (...): Adjust calls to ix86_vec_cost.
1248 2018-10-18  Richard Biener  <rguenther@suse.de>
1250         PR middle-end/87087
1251         Revert
1252         2018-02-07  Richard Biener  <rguenther@suse.de>
1254         PR tree-optimization/84204
1255         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1256         this place.
1258 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
1260         PR target/87537
1261         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
1262         of vec_duplicate.
1263         (test_vector_ops_duplicate): Add test for a scalar subreg of a
1264         VEC_MERGE of a VEC_DUPLICATE.
1266 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
1268         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
1269         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
1270         * doc/standards.texi (C Language): Document C2X.
1271         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
1272         (rl78_option_override): Handle "GNU C2X" language name.
1274 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
1276         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
1277         Document C17 as published in 2018.
1279 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
1281         PR middle-end/87623
1282         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
1283         bail out if both sides do not have the same storage order.
1285 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
1287         * bitmap.c (bitmap_head::dump): New.
1288         * bitmap.h (bitmap_head): Add dump().
1289         * gimple-ssa-evrp-analyze.c
1290         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
1291         (evrp_range_analyzer::set_ssa_range_info): Same.
1292         (evrp_range_analyzer::record_ranges_from_phis): Same.
1293         (evrp_range_analyzer::record_ranges_from_stmt): Same.
1294         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
1295         * gimple-ssa-sprintf.c (get_int_range): Same.
1296         (format_integer): Same.
1297         (sprintf_dom_walker::handle_gimple_call): Same.
1298         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
1299         (ipcp_vr_lattice::top_p): Same.
1300         (ipcp_vr_lattice::bottom_p): Same.
1301         (ipcp_vr_lattice::set_to_bottom): Same.
1302         (ipa_vr_operation_and_type_effects): Same.
1303         (propagate_vr_across_jump_function): Same.
1304         (ipcp_store_vr_results): Same.
1305         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
1306         (ipa_print_node_jump_functions_for_edge): Same.
1307         (ipa_get_value_range): Same.
1308         (ipa_compute_jump_functions_for_edge): Same.
1309         (ipa_write_jump_function): Same.
1310         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
1311         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
1312         Same.
1313         * vr-values.c (set_value_range_to_nonnegative): Same.
1314         (set_value_range_to_truthvalue): Same.
1315         (vr_values::get_value_range): Same.
1316         (vr_values::set_defs_to_varying): Same.
1317         (vr_values::update_value_range): Same.
1318         (symbolic_range_based_on_p): Same.
1319         (vr_values::op_with_boolean_value_range_p): Same.
1320         (vr_values::extract_range_for_var_from_comparison_expr): Same.
1321         (vr_values::extract_range_from_ssa_name): Same.
1322         (vr_values::extract_range_from_binary_expr): Same.
1323         (vr_values::extract_range_from_unary_expr): Same.
1324         (vr_values::extract_range_from_cond_expr): Same.
1325         (vr_values::extract_range_from_comparison): Same.
1326         (vr_values::check_for_binary_op_overflow): Same.
1327         (vr_values::extract_range_basic): Same.
1328         (vr_values::extract_range_from_assignment): Same.
1329         (compare_ranges): Same.
1330         (compare_range_with_value): Same.
1331         (vr_values::adjust_range_with_scev): Same.
1332         (vrp_valueize): Same.
1333         (vrp_valueize_1): Same.
1334         (vr_values::get_vr_for_comparison): Same.
1335         (vr_values::compare_name_with_value): Same.
1336         (vr_values::compare_names): Same.
1337         (vr_values::vrp_evaluate_conditional): Same.
1338         (find_case_label_ranges): Same.
1339         (vr_values::vrp_visit_switch_stmt): Same.
1340         (vr_values::extract_range_from_phi_node): Same.
1341         (vr_values::simplify_div_or_mod_using_ranges): Same.
1342         (vr_values::simplify_bit_ops_using_ranges): Same.
1343         (test_for_singularity): Same.
1344         (range_fits_type_p): Same.
1345         (vr_values::simplify_cond_using_ranges_1): Same.
1346         (vr_values::simplify_switch_using_ranges): Same.
1347         (vr_values::simplify_float_conversion_using_ranges): Same.
1348         (vr_values::two_valued_val_range_p): Same.
1349         (vr_values::add_equivalence): Move to value_range::equiv_add.
1350         * vr-values.h (vr_values::add_equivalence): Remove.
1351         (VR_INITIALIZER): Remove.
1352         * tree-vrp.c (value_range::set): New.
1353         (value_range::equiv_add): New.
1354         (value_range::value_range): New.
1355         (value_range::deep_copy): New.
1356         (value_range::check): New.
1357         (value_range::equal_p): New.
1358         (value_range::ignore_equivs_equal_p): New.
1359         (value_range::operator==): New.
1360         (value_range::operator!=): New.
1361         (value_range::symbolic_p): New.
1362         (value_range::numeric_p): New.
1363         (value_range::set_undefined): New.
1364         (value_range::set_varying): New.
1365         (value_range::may_contain_p): New.
1366         (value_range::equiv_clear): New.
1367         (value_range::singleton_p): New.
1368         (value_range::intersect): New.
1369         (value_range::dump): New.
1370         (value_range::set_and_canonicalize): New.
1371         (set_value_range): Adjust for value_range API.
1372         (set_value_range_to_undefined): Same.
1373         (set_value_range_to_varying): Same.
1374         (set_and_canonicalize_value_range): Same.
1375         (set_value_range_to_nonnull): Same.
1376         (set_value_range_to_null): Same.
1377         (range_is_null): Same.
1378         (range_is_nonnull): Same.
1379         (range_int_cst_p): Same.
1380         (range_int_cst_singleton_p): Same.
1381         (symbolic_range_p): Same.
1382         (range_includes_zero_p): Same.
1383         (value_range_constant_singleton): Same.
1384         (vrp_set_zero_nonzero_bits): Same.
1385         (ranges_from_anti_range): Same.
1386         (extract_range_into_wide_ints): Same.
1387         (extract_range_from_multiplicative_op): Same.
1388         (set_value_range_with_overflow): Same.
1389         (extract_range_from_binary_expr_1): Same.
1390         (extract_range_from_unary_expr): Same.
1391         (dump_value_range): Same.
1392         (debug_value_range): Same.
1393         (vrp_prop::check_array_ref): Same.
1394         (vrp_prop::check_mem_ref): Same.
1395         (vrp_prop::vrp_initialize): Same.
1396         (vrp_prop::visit_stmt): Same.
1397         (intersect_ranges): Same.
1398         (vrp_prop::visit_phi): Same.
1399         (vrp_prop::vrp_finalize): Same.
1400         (determine_value_range_1): Same.
1401         (determine_value_range): Same.
1402         (vrp_intersect_ranges_1): Rename to...
1403         (vrp_intersect_1): this.
1404         (vrp_intersect_ranges): Rename to...
1405         (value_range::intersect_helper): ...this.
1406         (vrp_meet_1): Rename to...
1407         (value_range::union_helper): ...this.
1408         (vrp_meet): Rename to...
1409         (value_range::union_): ...this.
1410         (copy_value_range): Remove.
1411         * tree-vrp.h (struct value_range): Rewrite into a proper class.
1412         (value_range::vrtype): New.
1413         (value_range::type): New.
1414         (value_range::equiv): New.
1415         (value_range::min): New.
1416         (value_range::max): New.
1417         (value_range::varying_p): New.
1418         (value_range::undefined_p): New.
1419         (value_range::null_p): New.
1420         (value_range::equiv_add): New.
1421         (copy_value_range): Remove.
1423 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
1425         * Makefile.in (SELFTEST_TARGETS): New.
1426         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
1427         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
1428         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
1429         c/Make-lang.in.
1430         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
1431         (selftest-c++-gdb, selftest-c++-valgrind): Move to
1432         cp/Make-lang.in.
1433         * configure: Regenerate.
1434         * configure.ac (selftest_languages): New.
1436 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
1438         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
1439         overflow wraps argument.
1440         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
1441         wide_int_range_multiplicative_op.
1442         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
1443         overflow wraps argument.
1444         (wide_int_range_multiplicative_op): Same.
1445         (wide_int_range_lshift): Same.
1446         (wide_int_range_div): Same.
1447         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
1448         (wide_int_range_lshift): Same.
1449         (wide_int_range_div): Same.
1451 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
1453         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
1454         use sign as argument.
1455         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
1456         wide_int_range_shift_undefined_p.
1458 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1460         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
1461         Rename to...
1462         (@despeculate_copy<ALLI_TI:mode>): ... This.
1463         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
1464         switch statement.
1466 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1468         * config.gcc: Obsolete *-*-solaris2.10*.
1469         * doc/install.texi (Specific, *-*-solaris2*): Document it.
1471 2018-10-12  Jeff Law  <law@redhat.com>
1473         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
1474         reg + sym +- const_int addressing modes.
1476 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
1478         * common.opt (fdiagnostics-minimum-margin-width=): New option.
1479         * diagnostic-show-locus.c (layout::layout): Apply the minimum
1480         margin width.
1481         (layout::start_annotation_line): Only print up to 3 of the
1482         margin character, to avoid touching the left-hand side.
1483         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
1484         minimum margin width, as set by test_diagnostic_context's ctor.
1485         (selftest::test_fixit_insert_containing_newline): Likewise.
1486         (selftest::test_fixit_insert_containing_newline_2): Likewise.
1487         (selftest::test_line_numbers_multiline_range): Clear
1488         dc.min_margin_width.
1489         * diagnostic.c (diagnostic_initialize): Initialize
1490         min_margin_width.
1491         * diagnostic.h (struct diagnostic_context): Add field
1492         "min_margin_width".
1493         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
1494         * opts.c (common_handle_option): Handle
1495         OPT_fdiagnostics_minimum_margin_width_.
1496         * selftest-diagnostic.c
1497         (selftest::test_diagnostic_context::test_diagnostic_context):
1498         Initialize min_margin_width to 6.
1499         * toplev.c (general_init): Initialize global_dc->min_margin_width.
1501 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
1503         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
1504         Fix usage of "error_at_rich_loc" in the comment.
1506 2018-10-15  Renlin Li  <renlin.li@arm.com>
1508         PR target/87563
1509         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
1510         if-conversioned loop when it contains ifn with types not
1511         supported by backend.
1512         * internal-fn.c (expand_direct_optab_fn): Add an assert.
1513         (direct_internal_fn_supported_p): New helper function.
1514         * internal-fn.h (direct_internal_fn_supported_p): Declare.
1516 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
1518         PR target/87572
1519         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
1520         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
1521         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
1523 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
1525         PR tree-optimization/87022
1526         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
1527         bits in dist vector rather than the first one.
1529 2018-10-15  Richard Biener  <rguenther@suse.de>
1531         PR middle-end/87610
1532         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
1533         (visit_loadstore): When a used restrict tag escaped verify that
1534         the points-to solution of "other" pointers do not include
1535         escaped.
1536         (compute_dependence_clique): If a used restrict tag escaped
1537         communicated that down to visit_loadstore.
1539 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
1541         * config/s390/s390.c (s390_expand_vec_init): Force vector element
1542         into reg if it isn't a general operand.
1544 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
1546         PR target/87599
1547         * config/i386/sse.md (*vec_dupv2di): Add register source to
1548         movddup.
1550 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
1552         PR target/87572
1553         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
1554         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
1555         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
1556         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
1558 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
1560         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
1561         (notice_args_size): Set it in the current trace if no insn that can
1562         throw internally has been seen yet.
1563         (connect_traces): When connecting args_size between traces, allow the
1564         incoming values not to match if there is an insn setting it before the
1565         first insn that can throw internally; in that case, force the creation
1566         of a CFI note on this latter insn.
1568 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
1570         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
1571         the base class, because using the injected-class-name was not clearly
1572         specified until DR 176.
1574 2018-10-12  Paul Koning  <ni1d@arrl.net>
1576         * config/pdp11/pdp11.md (doloop_end): New expander.
1577         (doloop_end_insn): renamed from "doloop_end".
1578         (addqi3): New pattern.
1579         (subqi3): New pattern.
1580         * config/pdp11/predicates.md (incdec_operand): New predicate.
1582 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
1584         PR middle-end/81376
1585         * real.c (format_helper::can_represent_integral_type_p): New function
1586         * real.h (format_helper::can_represent_integral_type_p): Ditto.
1587         * match.pd: New pattern.
1589 2018-10-12  Alexandre Oliva <oliva@adacore.com>
1591         * configure.ac: Introduce --enable-large-address-aware
1592         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
1593         * doc/install.texi: Document it.
1594         * configure, config.in: Rebuilt.
1595         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
1596         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
1597         (LINK_SPEC): Insert it.
1598         * config/i386/mingw-264.h: Likewise.
1600         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
1602 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
1604         PR rtl-optimization/87600
1605         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
1607 2018-10-12  Paul Koning  <ni1d@arrl.net>
1609         * doc/md.texi (doloop_end): Document that the pattern code may
1610         need to check operand mode.
1612 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
1614         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
1615         to zero-extend between int and floating-point registers.
1616         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
1617         ldp into floating-point registers.  Add type and arch attributes.
1618         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
1619         Use f_loads for type attribute.
1621 2018-10-11  Martin Sebor  <msebor@redhat.com>
1623         * doc/extend.texi (attribute packed): Correct typos.
1625 2018-10-11  Martin Sebor  <msebor@redhat.com>
1627         * doc/extend.texi (attribute flatten): Mention interaction with
1628         noinline.
1630 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
1632         PR target/87156
1633         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
1634         Set new_decl virtual flag to zero.
1636 2018-10-11  Martin Sebor  <msebor@redhat.com>
1638         PR middle-end/87593
1639         * doc/extend.texi (attribute format_arg): Discuss using multiple
1640         attributes on a single function.
1642 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
1644         PR tree-optimization/86829
1645         * match.pd (sin (atan (x))): New simplification rules.
1646         (cos (atan (x))): Likewise.
1647         * real.c (build_sinatan_real): New function.
1648         * real.h (build_sinatan_real): Prototype.
1650 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
1652         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
1653         function.
1654         (fold_mergeeo_helper): New helper function.
1655         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
1656         intrinsics.  Correct some whitespace indentation issues.
1658 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
1660         PR target/87511
1661         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
1662         Use HOST_WIDE_INT_1U for shift.
1664 2018-10-11  Doug Rupp  <rupp@adacore.com>
1665             Olivier Hainque  <hainque@adacore.com>
1667         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
1668         Pass --relax to the linker for RTPs.
1669         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
1671 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
1672             Jan Hubicka  <jh@suse.cz>
1673             Martin Jambor  <mjambor@suse.cz>
1675         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
1676         the same elements are repeated rather than printing all of them.
1677         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
1678         "repeated" elements.
1679         * read-rtl-function.c (test_loading_repeat): New function.
1680         (read_rtl_function_c_tests): Call test_loading_repeat.
1681         * rtl-tests.c (test_dumping_repeat): New function.
1682         (rtl_tests_c_tests): Call test_dumping_repeat.
1684 2018-10-11  Richard Biener  <rguenther@suse.de>
1686         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
1687         bdver?_cost): Unify to ...
1688         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
1689         * config/i386/i386.c (processor_cost_table): Adjust.
1691 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
1693         PR middle-end/87574
1694         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
1695         the thunk when expanding to GIMPLE.
1697 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1699         * varasm.c (mergeable_string_section): Don't try to move zero-length
1700         strings to the merge section.
1702 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
1704         PR target/87573
1705         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
1707 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
1709         PR target/87550
1710         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
1711         to special_args set.
1713 2018-10-10  Richard Biener  <rguenther@suse.de>
1715         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
1716         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
1717         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
1718         and recursing and pattern terminating the recursion on SSE
1719         vector width using ix86_expand_reduc.
1720         (reduc_sminmax_scal_<mode>): Split into part reducing to half
1721         width and recursing and SSE2 vector variant doing the final
1722         reduction with ix86_expand_reduc.
1723         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
1724         with terminating the recursion at AVX level, splitting that
1725         to SSE there.
1727 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
1729         * genmatch.c (error_cb): Rename to...
1730         (diagnostic_cb): ...this, converting int params to enums.
1731         (fatal_at): Update for renaming.
1732         (warning_at): Likewise.
1733         (main): Likewise.
1734         * input.c (selftest::ebcdic_execution_charset::apply):
1735         Update for renaming of...
1736         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
1737         to...
1738         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
1739         converting level and reason to enums.
1740         (class selftest::lexer_error_sink): Rename to...
1741         (class selftest::lexer_test_options): ...this, renaming field
1742         "m_errors" to "m_diagnostics".
1743         (selftest::lexer_test_options::apply): Update for renaming of...
1744         (selftest::lexer_test_options::on_error): ...this, renaming to...
1745         (selftest::lexer_test_options::on_diagnostic): ...this
1746         converting level and reason to enums.
1747         (selftest::test_lexer_string_locations_raw_string_unterminated):
1748         Update for renamings.
1749         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
1750         "reason".
1752 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
1754         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
1755         * config/rs6000/pmmintrin.h: New file.
1757 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
1759         PR tree-optimization/86659
1760         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
1762 2018-10-09  Richard Biener  <rguenther@suse.de>
1764         PR tree-optimization/63155
1765         * tree-ssa-structalias.c: Include tree-ssa.h.
1766         (get_constraint_for_ssa_var): For undefs return nothing_id.
1767         (find_func_aliases): Cleanup PHI handling.
1769 2018-10-09  Richard Biener  <rguenther@suse.de>
1771         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
1772         replacements.
1774 2018-10-09  Martin Liska  <mliska@suse.cz>
1776         * asan.c (asan_emit_stack_protection): If a stack variable
1777         is located in a same file as current function, then emit
1778         line info into variable definition string.
1780 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
1782         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
1783         information.
1785 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
1787         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
1788         on the thunk.
1790 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
1792         PR tree-optimization/86659
1793         * gimple-match.h (struct gimple_match_op): Add reverse field.
1794         (gimple_match_op::set_op): New overloaded method.
1795         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
1796         the REF_REVERSE_STORAGE_ORDER flag on the value.
1797         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
1798         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
1799         
1800 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
1802         PR middle-end/63155
1803         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
1804         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
1806 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
1808         PR target/87517
1809         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
1810         Defined with __builtin_ia32_vfmaddsubpd512_mask.
1812 2018-10-08  Richard Biener  <rguenther@suse.de>
1814         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
1815         cost the same as AVX128 ones.
1817 2018-10-08  Paul Koning  <ni1d@arrl.net>
1819         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
1820         (expand_block_move): New function.
1821         * config/pdp11/pdp11.c (output_block_move): Remove.
1822         (expand_block_move): New function.
1823         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
1824         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
1825         (*movmemhi1): Remove.
1827 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
1829         * config/s390/2827.md: Increase latencies for some FP instructions.
1831 2018-10-08  Richard Biener  <rguenther@suse.de>
1833         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
1834         Open a dump scope.
1835         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
1836         * tree-vectorizer.h (dump_stmt_cost): Adjust.
1837         (add_stmt_cost): Dump return value of the hook.
1839 2018-10-08  Richard Biener  <rguenther@suse.de>
1841         PR tree-optimization/63155
1842         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
1843         (ssa_propagation_engine::ssa_propagate): Remove redundant
1844         bitmap bit clearing.
1846 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
1848         PR rtl-optimization/86939
1849         PR rtl-optimization/87479
1850         * ira.h (non_conflicting_reg_copy_p): New prototype.
1851         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
1852         (make_hard_regno_dead): Don't add conflicts for register
1853         ignore_reg_for_conflicts.
1854         (make_object_dead): Likewise.
1855         (non_conflicting_reg_copy_p): New function.
1856         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
1857         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
1858         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
1859         (make_hard_regno_dead): Don't add conflicts for register
1860         ignore_reg_for_conflicts.  Remove special conflict handling of
1861         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
1862         check_pic_pseudo_p and update callers.
1863         (mark_pseudo_dead): Don't add conflicts for register
1864         ignore_reg_for_conflicts.
1865         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
1867 2018-10-05  Andrew Waterman  <andrew@sifive.com>
1868             Jim Wilson  <jimw@sifive.com>
1870         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
1871         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
1872         new pattern using HONOR_SNANS that emits one extra instruction.
1874 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
1876         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
1877         patterns): Merge SI and DI patterns to a GPR pattern.
1878         (unnamed define_insn and define_split for record form of that): Merge
1879         to a single define_insn_and_split pattern.
1881 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
1883         PR c++/56856
1884         * input.c (expand_location_to_spelling_point): Add param "aspect"
1885         and use rather than hardcoding LOCATION_ASPECT_CARET.
1886         (get_substring_ranges_for_loc): Handle the case of a single token
1887         within a macro expansion.
1888         * input.h (expand_location_to_spelling_point): Add "aspect" param,
1889         defaulting to LOCATION_ASPECT_CARET.
1891 2018-10-05  Paul Koning  <ni1d@arrl.net>
1893         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
1894         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
1895         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
1896         (pdp11_guard_type): New function.
1898 2018-10-05  Paul Koning  <ni1d@arrl.net>
1900         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
1901         * config/pdp11/pdp11.opt (mfloat32): Remove.
1902         (mfloat64): Remove.
1903         * doc/invoke.texi (pdp11 -mfloat32): Remove:
1904         (pdp11 -mfloat64): Remove.
1906 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
1908         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
1909         (*cmp<mode>_cc_i387): Ditto.
1910         (*cmpu<mode>_cc_i387): Ditto.
1911         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
1912         * config/i386/i386.c (ix86_expand_fp_compare): Remove
1913         "scratch" argument.
1914         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
1915         Emit x86_sahf_1 pattern.
1916         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
1917         (ix86_expand_carry_flag_compare): Ditto.
1919 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
1921         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
1922         to reg_or_0_operand.  Add "C" constraint.
1923         (*cmpxf_cc_i387): Ditto.
1924         (*cmp<mode>_i387): Change operand 2 predicate
1925         to nonimm_or_0_operand.  Add "C" constraint.
1926         (*cmp<mode>_cc_i387): Ditto.
1927         (*cmp<mode>_0_i387): Remove insn pattern.
1928         (*cmp<mode>_0_cc_i387): Ditto.
1930 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
1932         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
1933         * config/i386/predicates.md (nonimm_or_0_operand): Rename
1934         from vector_move_operand.  Update all uses.
1936 2018-10-05  Martin Sebor  <msebor@redhat.com>
1938         PR tree-optimization/87490
1939         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
1940         consistently.
1942 2018-10-05  Richard Biener  <rguenther@suse.de>
1944         PR tree-optimization/63155
1945         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
1946         vertical space in dumpfiles.
1947         * tree-ssa-propagate.h
1948         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
1949         * tree-ssa-propagate.c (cfg_blocks_back): New global.
1950         (ssa_edge_worklist_back): Likewise.
1951         (curr_order): Likewise.
1952         (cfg_blocks_get): Remove abstraction.
1953         (cfg_blocks_add): Likewise.
1954         (cfg_blocks_empty_p): Likewise.
1955         (add_ssa_edge): Add to current or next worklist based on
1956         RPO index.
1957         (add_control_edge): Likewise.
1958         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
1959         into ...
1960         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
1961         iteration from CFG and SSA edge worklist so we process
1962         everything in RPO order, prioritizing forward progress
1963         over iteration.
1964         (ssa_prop_init): Allocate new worklists, do not dump
1965         immediate uses.
1966         (ssa_prop_fini): Free new worklists.
1968 2018-10-05  Richard Biener  <rguenther@suse.de>
1970         * tree-core.h (tree_block::abstract_flag): Remove.
1971         (tree_block::block_num): Make full 32bits.
1972         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
1973         * tree.h (BLOCK_ABSTRACT): Remove.
1974         * dwarf2out.c (gen_lexical_block_die): Remove dead code
1975         resulting from BLOCK_ABSTRACT being always false.
1976         (gen_inlined_subroutine_die): Likewise.
1977         (gen_block_die): Likewise.
1978         * tree.c (block_ultimate_origin): Likewise.
1979         * tree-pretty-print.c (dump_block_node): Remove code dealing
1980         with BLOCK_ABSTRACT.
1981         * tree-ssa-live.c (dump_scope_block): Likewise.
1982         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
1983         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
1985 2018-10-05   Richard Biener  <rguenther@suse.de>
1987         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
1988         is asked for initialize mode to the component mode of the
1989         vector type.
1991 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
1993         PR target/87522
1994         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
1995         assembler for -mavx.
1996         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
1998 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
2000         PR target/87509
2001         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
2002         RS6000_BTM_DFP.
2003         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
2004         to be DImode.  When using mffscrn, force the operand to a register.
2006 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
2008         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
2009         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
2010         X87MODEF mode iterator.
2011         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
2012         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
2013         X87MODEF mode iterator.
2015 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
2017         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
2018         -Wno-prio-ctor-dtor.
2020 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2022         * Makefile.in (OBJS): Add opt-problem.o.
2023         * dump-context.h: Include "selftest.h.
2024         (selftest::temp_dump_context): New forward decl.
2025         (class dump_context): Make friend of class
2026         selftest::temp_dump_context.
2027         (dump_context::dump_loc_immediate): New decl.
2028         (class dump_pretty_printer): Move here from dumpfile.c.
2029         (class temp_dump_context): Move to namespace selftest.
2030         (temp_dump_context::temp_dump_context): Add param
2031         "forcibly_enable_dumping".
2032         (selftest::verify_dumped_text):
2033         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
2034         (selftest::verify_item):
2035         (ASSERT_IS_TEXT): Move here from dumpfile.c.
2036         (ASSERT_IS_TREE): Likewise.
2037         (ASSERT_IS_GIMPLE): Likewise.
2038         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
2039         to...
2040         (dump_context::dump_loc_immediate): ...this new function.
2041         (class dump_pretty_printer): Move to dump-context.h.
2042         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
2043         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
2044         (temp_dump_context::temp_dump_context): Move to "selftest"
2045         namespace.  Add param "forcibly_enable_dumping", and use it to
2046         conditionalize the use of m_pp;
2047         (selftest::verify_dumped_text): Make non-static.
2048         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
2049         (selftest::verify_item): Make non-static.
2050         (ASSERT_IS_TEXT): Move to dump-context.h.
2051         (ASSERT_IS_TREE): Likewise.
2052         (ASSERT_IS_GIMPLE): Likewise.
2053         (selftest::test_capture_of_dump_calls): Pass "true" for new
2054         param of temp_dump_context.
2055         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
2056         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
2057         TDF_COMPARE_DEBUG.
2058         * opt-problem.cc: New file.
2059         * opt-problem.h: New file.
2060         * optinfo-emit-json.cc
2061         (selftest::test_building_json_from_dump_calls): Pass "true" for
2062         new param of temp_dump_context.
2063         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
2064         (optinfo::emit_for_opt_problem): New function.
2065         (optinfo::emit): Clarity which emit_item is used.
2066         * optinfo.h (optinfo::get_dump_location): New accessor.
2067         (optinfo::emit_for_opt_problem): New decl.
2068         (optinfo::emit): Make const.
2069         * selftest-run-tests.c (selftest::run_tests): Call
2070         selftest::opt_problem_cc_tests.
2071         * selftest.h (selftest::opt_problem_cc_tests): New decl.
2072         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
2073         bool to opt_result, converting fprintf messages to
2074         opt_result::failure_at calls.  Add "stmt" param for use by the
2075         failure_at calls.
2076         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
2077         (runtime_alias_check_p): Convert return type from bool to
2078         opt_result, converting dump_printf calls to
2079         opt_result::failure_at, using the statement DDR_A for their
2080         location.
2081         (find_data_references_in_stmt): Convert return type from bool to
2082         opt_result, converting "return false" to opt_result::failure_at
2083         with a new message.
2084         * tree-data-ref.h: Include "opt-problem.h".
2085         (dr_analyze_innermost): Convert return type from bool to opt_result,
2086         and add a const gimple * param.
2087         (find_data_references_in_stmt): Convert return type from bool to
2088         opt_result.
2089         (runtime_alias_check_p): Likewise.
2090         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
2091         dr_analyze_innermost.
2092         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
2093         Convert return type from bool to opt_result, adding a message for
2094         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
2095         (vect_analyze_data_ref_dependence): Convert return type from bool
2096         to opt_result.  Change sense of return type from "false"
2097         effectively meaning "no problems" to "false" meaning a problem,
2098         so that "return false" becomes "return opt_result::success".
2099         Convert "return true" calls to opt_result::failure_at, using
2100         the location of statement A rather than vect_location.
2101         (vect_analyze_data_ref_dependences): Convert return type from bool
2102         to opt_result.
2103         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
2104         calls to opt_result::failure_at, using the stmt location rather
2105         than vect_location.
2106         (vect_verify_datarefs_alignment): Convert return type from bool
2107         to opt_result.
2108         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
2109         into multiple more-tightly-scoped copies.
2110         (vect_analyze_data_refs_alignment): Convert return type from bool
2111         to opt_result.
2112         (vect_analyze_data_ref_accesses): Likewise, converting a
2113         "return false" to a "return opt_result::failure_at", adding a
2114         new message.
2115         (vect_prune_runtime_alias_test_list): Convert return type from
2116         bool to opt_result, converting dump_printf_loc to
2117         opt_result::failure_at.  Add a %G to show the pertinent statement,
2118         and use the stmt's location rather than vect_location.
2119         (vect_find_stmt_data_reference): Convert return type from
2120         bool to opt_result, converting dump_printf_loc to
2121         opt_result::failure_at, using stmt's location.
2122         (vect_analyze_data_refs):  Convert return type from bool to
2123         opt_result.  Convert "return false" to "return
2124         opt_result::failure_at", adding messages as needed.
2125         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
2126         type from bool to opt_result.
2127         (vect_determine_vf_for_stmt): Likewise.
2128         (vect_determine_vectorization_factor): Likewise, converting
2129         dump_printf_loc to opt_result::failure_at, using location of phi
2130         rather than vect_location.
2131         (vect_analyze_loop_form_1): Convert return type from bool to
2132         opt_result, converting dump_printf_loc calls, retaining the use of
2133         vect_location.
2134         (vect_analyze_loop_form): Convert return type from loop_vec_info
2135         to opt_loop_vec_info.
2136         (vect_analyze_loop_operations): Convert return type from bool to
2137         opt_result, converting dump_printf_loc calls, using the location
2138         of phi/stmt rather than vect_location where available.  Convert
2139         various "return false" to "return opt_result::failure_at" with
2140         "unsupported phi" messages.
2141         (vect_get_datarefs_in_loop): Convert return type from bool to
2142         opt_result.  Add a message for the
2143         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
2144         (vect_analyze_loop_2): Convert return type from bool to
2145         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
2146         each "goto again;", adding new messages where needed.
2147         Add "unsupported grouped {store|load}" messages.
2148         (vect_analyze_loop): Convert return type from loop_vec_info to
2149         opt_loop_vec_info.
2150         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
2151         bool to opt_result.
2152         * tree-vect-stmts.c (process_use): Likewise, converting
2153         dump_printf_loc call and using stmt location, rather than
2154         vect_location.
2155         (vect_mark_stmts_to_be_vectorized): Likeise.
2156         (vect_analyze_stmt): Likewise, adding a %G.
2157         (vect_get_vector_types_for_stmt): Convert return type from bool to
2158         opt_result, converting dump_printf_loc calls and using stmt
2159         location, rather than vect_location.
2160         (vect_get_mask_type_for_stmt): Convert return type from tree to
2161         opt_tree, converting dump_printf_loc calls and using stmt location.
2162         * tree-vectorizer.c: Include "opt-problem.h.
2163         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
2164         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
2165         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
2166         enabled, use it to report at the top level "couldn't vectorize
2167         loop" followed by the problem.
2168         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
2169         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
2170         to opt_result.
2171         (vect_analyze_stmt): Likewise.
2172         (vect_get_vector_types_for_stmt): Likewise.
2173         (tree vect_get_mask_type_for_stmt): Likewise.
2174         (vect_analyze_data_ref_dependences): Likewise.
2175         (vect_enhance_data_refs_alignment): Likewise.
2176         (vect_analyze_data_refs_alignment): Likewise.
2177         (vect_verify_datarefs_alignment): Likewise.
2178         (vect_analyze_data_ref_accesses): Likewise.
2179         (vect_prune_runtime_alias_test_list): Likewise.
2180         (vect_find_stmt_data_reference): Likewise.
2181         (vect_analyze_data_refs): Likewise.
2182         (vect_analyze_loop): Convert return type from loop_vec_info to
2183         opt_loop_vec_info.
2184         (vect_analyze_loop_form): Likewise.
2185         (vect_analyze_slp): Convert return type from bool to opt_result.
2187 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2189         * doc/invoke.texi (-fopt-info): Document new "internals"
2190         sub-option.
2191         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
2192         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
2193         MSG_ALL_KINDS.
2194         (optinfo_verbosity_options): Add "internals".
2195         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
2196         (dump_context::apply_dump_filter_p): New member function.
2197         (dump_context::dump_loc): Use apply_dump_filter_p rather than
2198         explicitly masking the dump_kind.
2199         (dump_context::begin_scope): Increment the scope depth first.  Use
2200         apply_dump_filter_p rather than explicitly masking the dump_kind.
2201         (dump_context::emit_item): Use apply_dump_filter_p rather than
2202         explicitly masking the dump_kind.
2203         (dump_dec): Likewise.
2204         (dump_hex): Likewise.
2205         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
2206         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
2207         (opt_info_switch_p): Update handling of default
2208         MSG_OPTIMIZED_LOCATIONS to cope with default of
2209         MSG_PRIORITY_USER_FACING.
2210         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
2211         masking the dump_kind.
2212         (selftest::test_capture_of_dump_calls): Update test_dump_context
2213         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
2214         than MSG_ALL.  Generalize scope test to be run at all four
2215         combinations of with/without MSG_PRIORITY_USER_FACING and
2216         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
2217         for each of the two values.
2218         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
2219         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
2220         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
2221         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
2222         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
2223         with MSG_PRIORITY_*.
2224         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
2225         dump messages as MSG_PRIORITY_USER_FACING.
2226         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
2227         about the interaction with MSG_PRIORITY_*.
2229 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2231         * varasm.c (output_constant): Add new parameter merge_strings.
2232         Make strings properly zero terminated in merge string sections.
2233         (mergeable_string_section): Don't fail if the last char is non-zero.
2234         (assemble_variable_contents): Handle merge string sections.
2235         (assemble_variable): Likewise.
2236         (assemble_constant_contents): Likewise.
2237         (output_constant_def_contents): Likewise.
2238         (output_constructor_array_range,
2239         output_constructor_regular_field): Adjust call to output_constant.
2240         (output_object_block): Adjust call to assemble_constant_contents
2241         and assemble_variable_contents.
2243 2018-10-04  Martin Liska  <mliska@suse.cz>
2245         PR c/87483
2246         * cgraphunit.c (process_function_and_variable_attributes):
2247         Warn about a function with alias attribute and a body.
2249 2018-10-04  Martin Liska  <mliska@suse.cz>
2251         PR ipa/82625
2252         * multiple_target.c (redirect_to_specific_clone): New function.
2253         (ipa_target_clone): Use it.
2254         * tree-inline.c: Fix comment.
2256 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
2258         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
2259         fields.
2260         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
2261         (gcc::dump_manager::register_pass): New member function, adapted
2262         from loop body in gcc::pass_manager::register_pass, adding a
2263         call to update_dfi_for_opt_info.
2264         (gcc::dump_manager::opt_info_enable_passes): Store the
2265         -fopt-info options into the new fields.  Move the loop
2266         bodies into...
2267         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
2268         function.
2269         * dumpfile.h (struct opt_pass): New forward decl.
2270         (gcc::dump_manager::register_pass): New decl.
2271         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
2272         (class gcc::dump_manager): Add fields "m_optgroup_flags",
2273         "m_optinfo_flags", and "m_optinfo_filename".
2274         * passes.c (gcc::pass_manager::register_pass): Move all of the
2275         dump-handling code to gcc::dump_manager::register_pass.
2277 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
2279         PR rtl-optimization/87466
2280         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
2281         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
2282         * doc/tm.texi: Regenerate.
2283         * ira-lives.c (process_bb_node_lives): Use the new target hook.
2284         * lra-lives.c (process_bb_lives): Likewise.
2285         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
2286         Define.
2288 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
2290         * params.c (add_params): Fix initialization.
2292 2018-10-04  Martin Liska  <mliska@suse.cz>
2294         PR gcov-profile/84107
2295         * tree-profile.c (init_ic_make_global_vars):
2296         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
2297         Come up with new ic_tuple* variables.  Emit
2298         __gcov_indirect_call{,_topn} variables.
2299         (gimple_gen_ic_profiler): Access the variable
2300         and emit gimple.
2301         (gimple_gen_ic_func_profiler): Access
2302         __gcov_indirect_call.callee field.
2303         (gimple_init_gcov_profiler): Use ptr_type_node.
2304         * value-prof.c (gimple_ic): Use ptr_type_node.
2306 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2308         PR tree-optimization/85787
2309         * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
2310         into this function and add support for detecting multiple phis.
2311         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
2313 2018-10-04  Martin Liska  <mliska@suse.cz>
2315         PR ipa/87491
2316         * ipa-inline.c (inline_to_all_callers_1):
2317         Call ultimate_alias_target for node being inlined.
2319 2018-10-03  Jeff Law  <law@redhat.com>
2321         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
2322         target's wchar_t.
2323         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
2324         * tree.h (get_typenode_from_name): Prototype.
2326 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
2328         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
2329         Change operand 2 predicate to nonimmediate_operand.
2330         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
2332 2018-10-03  Martin Sebor  <msebor@redhat.com>
2333             Jeff Law  <law@redhat.com>
2335         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
2336         initialize it.
2337         (get_string_length): Detect unterminated arrays.
2338         (format_string): Same.
2339         (format_directive): Warn about unterminated arrays.
2340         (handle_gimple_call): Mark statements with no_warning as needed.
2342 2018-10-03  Jim Wilson  <jimw@sifive.com>
2344         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
2345         also define __riscv_abi_rve.  Delete trailing white space.
2347 2018-10-03  Paul Koning  <ni1d@arrl.net>
2349         Enable LRA register allocator for PDP11.
2350         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
2351         (R): Likewise.
2352         (D): Likewise.
2353         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
2354         * config/pdp11/pdp11.opt (-mlra): New option.
2355         * doc/invoke.texi (PDP-11 Options): Document -mlra.
2357 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
2359         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
2360         (*<absneg:code>extend<mode>xf2): Ditto.
2362 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
2364         PR tree-optimization/87415
2365         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
2366         precision fields.
2368 2018-10-02  Jeff Law  <law@redhat.com>
2370         * gimple-fold.c (get_range_strlen): Only set *nonstr when
2371         an unterminated string is discovered.  Bubble up range
2372         even for unterminated strings.
2373         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
2374         indicates the string was not terminated via NONSTR.
2376 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
2378         * tree-vrp.c (extract_range_from_unary_expr): Special case all
2379         pointer conversions.
2380         Do not do anything special for anti-ranges.
2382 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
2384         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
2385         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
2387 2018-10-03  Martin Liska  <mliska@suse.cz>
2389         PR gcov-profile/86109
2390         * coverage.c (coverage_begin_function): Do not
2391         mark lambdas as artificial.
2392         * tree-core.h (struct GTY): Remove tm_clone_flag
2393         and introduce new lambda_function.
2394         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
2396 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
2398         PR target/87474
2399         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
2400         P8_VECTOR and VSX are enabled.
2402 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
2404         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
2405         0x3907 as CPU model number.
2407 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
2409         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
2410         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
2411         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
2412         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
2413         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
2414         * config/s390/s390.md: Likewise. Rename also the cpu attribute
2415         value from arch12 to z14.
2417 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
2419         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
2420         (isinfxf2): Ditto.
2421         (isinf<mode>2): Ditto.
2423 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
2425         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
2426         before emitting fxam.  Perform calculations in XFmode.
2428 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
2430         * match.pd (((X /[ex] A) +- B) * A): New transformation.
2432 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
2434         PR middle-end/87319
2435         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
2436         * tree.c (signed_or_unsigned_type_for): Handle complex.
2438 2018-10-02  Jeff Law  <law@redhat.com>
2440         * gimple-fold.c (get_range_strlen): Remove dead code.
2442 2018-10-02  Martin Sebor  <msebor@redhat.com>
2443             Jeff Law  <law@redhat.com>
2445         * builtins.c (unterminated_array): Add new arguments.
2446         If argument is not terminated, bubble up size and exact
2447         state to callers.
2448         (expand_builtin_strnlen): Detect, avoid expanding
2449         and diagnose unterminated arrays.
2450         (c_strlen): Fill in offset of start of unterminated strings.
2451         * builtins.h (unterminated_array): Update prototype.
2453 2018-10-02  Richard Biener  <rguenther@suse.de>
2455         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
2456         of haddv4df, first reduce to SSE width and exploit the fact
2457         that we only need element zero with the reduction result.
2458         (reduc_plus_scal_v2df): Likewise.
2460 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
2462         * dojump.h (do_jump): Delete.
2463         (do_jump_1): Likewise.
2464         (split_comparison): Move around.
2465         * dojump.c (do_jump): Make static.
2466         (do_jump_1): Likewise.
2467         (jumpifnot): Move around.
2468         (jumpifnot_1): Likewise.
2469         (jumpif): Likewise.
2470         (jumpif_1): Likewise.
2471         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
2473 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
2475         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
2476         insns in the delay slot and add_insn_after for the jump insn.
2478 2018-10-02  Richard Biener  <rguenther@suse.de>
2480         * tree-inline.c (expand_call_inline): Use the location of
2481         the callee declaration for the inline-entry marker.
2482         * final.c (notice_source_line): Remove special-casing of
2483         NOTE_INSN_INLINE_ENTRY.
2485 2018-10-01  Carl Love  <cel@us.ibm.com>
2487         PR 69431
2488         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
2489         (__builtin_mtfsb0): New.
2490         (__builtin_mtfsb1): New.
2491         ( __builtin_set_fpscr_rn): New.
2492         (__builtin_set_fpscr_drn): New.
2493         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
2494         (rs6000_expand_set_fpscr_rn_builtin): Add.
2495         (rs6000_expand_set_fpscr_drn_builtin): Add.
2496         (rs6000_expand_builtin): Add case statement entries for
2497         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
2498         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
2499         RS6000_BUILTIN_MFFSL.
2500         (rs6000_init_builtins): Add ftype initialization and def_builtin
2501         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
2502         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
2503         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
2504         rs6000_mffscdrn): Add define_insn.
2505         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
2506         * doc/extend.texi: Add documentation for the builtins.
2508 2018-10-01  Richard Biener  <rguenther@suse.de>
2510         PR tree-optimization/87465
2511         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
2512         causing branch miscounts.
2514 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2516         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
2517         aarch64_option_default_param):  New.
2518         (params.h): Include.
2519         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
2520         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
2521         stack-clash protection validation code.
2523 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2525         * params.c (validate_param): New.
2526         (add_params): Use it.
2527         (set_param_value): Refactor param validation into validate_param.
2528         (diagnostic.h): Include.
2529         * diagnostic.h (diagnostic_ready_p): New.
2531 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2533         * params.c (set_param_value):
2534         Add index of parameter being validated.
2535         * common/common-target.def (option_validate_param): New.
2536         * common/common-targhooks.h (default_option_validate_param): New.
2537         * common/common-targhooks.c (default_option_validate_param): New.
2538         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
2539         * doc/tm.texi: Regenerate.
2541 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2543         PR target/86486
2544         * config/aarch64/aarch64.c (aarch64_override_options_internal):
2545         Add validation for stack-clash parameters and set defaults.
2547 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2549         PR target/86486
2550         * configure.ac: Add stack-clash-protection-guard-size.
2551         * doc/install.texi: Document it.
2552         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
2553         * params.def: Update comment for guard-size.
2554         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
2555         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
2556         * configure: Regenerate.
2558 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2560         PR target/86486
2561         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
2562         STACK_DYNAMIC_OFFSET): New.
2563         * config/aarch64/aarch64.c (aarch64_layout_frame):
2564         Update outgoing args size.
2565         (aarch64_stack_clash_protection_alloca_probe_range,
2566         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
2568 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2570         PR target/86486
2571         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
2572         probe ranges.
2573         * target.def (stack_clash_protection_alloca_probe_range): New.
2574         (stack_clash_protection_final_dynamic_probe): Remove.
2575         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
2576         (default_stack_clash_protection_final_dynamic_probe): Remove.
2577         * targhooks.c: Likewise.
2578         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
2579         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
2580         * doc/tm.texi: Regenerate.
2582 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2584         PR target/86486
2585         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
2586         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
2587         aarch64_clamp_to_uimm12_shift): New.
2588         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
2589         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
2591 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
2593         PR target/86486
2594         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
2596 2018-10-01  Jeff Law  <law@redhat.com>
2597             Richard Sandiford <richard.sandiford@linaro.org>
2598             Tamar Christina  <tamar.christina@arm.com>
2600         PR target/86486
2601         * config/aarch64/aarch64.md
2602         (probe_stack_range): Add k (SP) constraint.
2603         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
2604         STACK_CLASH_MAX_UNROLL_PAGES): New.
2605         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
2606         stack probes for stack clash.
2607         (aarch64_allocate_and_probe_stack_space): New.
2608         (aarch64_expand_prologue): Use it.
2609         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
2610         (aarch64_sub_sp): Add emit_move_imm optional param.
2612 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
2614         PR tree-optimization/87261
2615         * match.pd: Remove trailing whitespace.
2616         Add (x & y) | ~(x | y) -> ~(x ^ y),
2617         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
2619 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
2621         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
2622         constraints.
2623         (ashlsi3_insn): Update instruction constraints.
2624         (ashrsi3_insn): Likewise.
2625         (rotrsi3): Likewise.
2626         (add_shift): Likewise.
2627         * config/arc/constraints.md (Csz): New 32 bit constraint. It
2628         avoids placing in the limm field small constants which, otherwise,
2629         could end into a small instruction.
2631 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
2633         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
2634         destination register is not odd-even.
2635         (umaddsidi4_split): Likewise.
2637 2018-10-01  Richard Biener  <rguenther@suse.de>
2639         * tree-inline.c (expand_call_inline): Store origin of fn
2640         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
2641         * tree.c (block_ultimate_origin): Simplify and do some
2642         checking.
2644 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
2646         * config/i386/mmx.md (EMMS): New int iterator.
2647         (emms): New int attribute.
2648         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
2649         EMMS int iterator.  Explicitly declare clobbers.
2650         (mmx_emms): Remove expander.
2651         (mmx_femms): Ditto.
2652         * config/i386/predicates.md (emms_operation): Remove predicate.
2653         (vzeroall_pattern): New predicate.
2654         (vzeroupper_pattern): Rename from vzeroupper_operation.
2655         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
2656         vzeroupper_pattern and vzeroall_pattern predicates.
2658 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
2660         PR rtl-optimization/86939
2661         * ira-lives.c (make_hard_regno_born): Rename from this...
2662         (make_hard_regno_live): ... to this.  Remove update to conflict
2663         information.  Update function comment.
2664         (make_hard_regno_dead): Add conflict information update.  Update
2665         function comment.
2666         (make_object_born): Rename from this...
2667         (make_object_live): ... to this.  Remove update to conflict information.
2668         Update function comment.
2669         (make_object_dead):  Add conflict information update.  Update function
2670         comment.
2671         (mark_pseudo_regno_live): Call make_object_live.
2672         (mark_pseudo_regno_subword_live): Likewise.
2673         (mark_hard_reg_dead): Update function comment.
2674         (mark_hard_reg_live): Call make_hard_regno_live.
2675         (process_bb_node_lives): Likewise.
2676         * lra-lives.c (make_hard_regno_born): Rename from this...
2677         (make_hard_regno_live): ... to this.  Remove update to conflict
2678         information.  Remove now uneeded check_pic_pseudo_p argument.
2679         Update function comment.
2680         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
2681         to conflict information.  Update function comment.
2682         (mark_pseudo_live): Remove update to conflict information.  Update
2683         function comment.
2684         (mark_pseudo_dead): Add conflict information update.
2685         (mark_regno_live): Call make_hard_regno_live.
2686         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
2687         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
2689 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
2691         PR target/87370
2692         * config/i386/i386.c (construct_container): Use TImode for
2693         BLKmode values in 2 integer registers.
2695 2018-09-29  Jeff Law  <law@redhat.com>
2697         * builtins.c (unterminated_array): Pass in c_strlen_data * to
2698         c_strlen rather than just a tree *.
2699         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
2700         Update recursive calls appropriately.  If caller did not provide a
2701         suitable data pointer, create a local one.  When a non-terminated
2702         string is discovered, bubble up information about the string via the
2703         c_strlen_data object.
2704         * builtins.h (c_strlen): Update prototype.
2705         (c_strlen_data): New structure.
2706         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
2707         For a type 2 call, if c_strlen indicates a non-terminated string
2708         use the length of the non-terminated string.
2709         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
2711 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
2713         PR target/87467
2714         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
2715         __m512d type for __A argument rather than __m512.
2717 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
2719         * match.pd (simple_comparison): Don't optimize if either operand is
2720         a function pointer when target needs function pointer canonicalization.
2722 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2724         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
2725         power5 .. power9 to remove indirection.
2726         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
2727         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
2728         ASM_CPU_476_SPEC): Delete.
2729         (ASM_CPU_SPEC): Adjust.
2730         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
2731         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
2733 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2735         * config.in: Delete HAVE_AS_DCI.
2736         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
2737         * config/rs6000/rs6000.h: Ditto.
2738         * configure.ac: Delete HAVE_AS_DCI.
2739         * configure: Regenerate.
2741 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
2743         * config.in (HAVE_AS_LWSYNC): Delete.
2744         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
2745         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
2746         do it as a .long .
2747         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
2748         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
2749         as a .long .
2750         * configure.ac: Delete HAVE_AS_LWSYNC.
2751         * configure: Regenerate.
2753 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
2754             Pierre-Marie de Rodat  <derodat@adacore.com>
2756         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
2757         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
2758         (cgraph_node::create_thunk): Add indirect_offset parameter.
2759         (thunk_adjust): Likewise.
2760         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
2761         and initialize the corresponding field with it.
2762         (cgraph_node::dump): Dump indirect_offset field.
2763         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
2764         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
2765         (thunk_adjust): Add indirect_offset parameter and deal with it.
2766         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
2767         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
2768         or if the thunk is external or local.  Fix formatting.  Do not chain
2769         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
2770         if any, in the GIMPLE representation.
2771         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
2772         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
2773         (input_node): Read indirect_offset field.
2774         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
2775         call to thunk_adjust.
2776         * tree-nested.c (struct nesting_info): Add thunk_p field.
2777         (create_nesting_tree): Set it.
2778         (convert_all_function_calls): Copy static chain from targets to thunks.
2779         (finalize_nesting_tree_1): Return early for thunks.
2780         (unnest_nesting_tree_1): Do not finalize thunks.
2781         (gimplify_all_functions): Do not gimplify thunks.
2783 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
2785         * opt-suggestions.c (option_proposer::build_option_suggestions):
2786         Release "option_values".
2788 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
2790         * coverage.c (get_coverage_counts): Convert problem-reporting dump
2791         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
2792         * dumpfile.c (kind_as_string): New function.
2793         (dump_loc): Rather than a hardcoded prefix of "note: ", use
2794         kind_as_string to vary the prefix based on dump_kind.
2795         (selftest::test_capture_of_dump_calls): Update for above.
2797 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
2799         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
2800         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
2802 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
2804         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
2805         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
2806         INVALID_REGNUM instead of FPSR_REG.
2807         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
2808         * config/i386/i386.md: Update comment of FP compares.
2809         (fldenv): Do not clobber FPSR_REG.
2811 2018-09-28  Richard Biener  <rguenther@suse.de>
2813         * tree.h (BLOCK_ORIGIN): New.
2814         * omp-expand.c (grid_expand_target_grid_body): Assign
2815         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
2816         * tree-inline.c (remap_block): Likewise.
2817         * auto-profile.c (get_function_decl_from_block): Simplify
2818         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
2819         * langhooks.c (lhd_print_error_function): Likewise.
2820         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
2821         Likewise.
2822         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
2823         * tree.c (block_nonartificial_location): Likewise.
2824         (block_ultimate_origin): Likewise.
2825         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
2826         no longer needed LTO case.
2828 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
2829             Jan Hubicka  <jh@suse.cz>
2830             Martin Jambor  <mjambor@suse.cz>
2832         * simplify-rtx.c (simplify_merge_mask): New function.
2833         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
2834         same masks are used in op1 or op2.
2835         (test_vec_merge): New function.
2836         (test_vector_ops): Call test_vec_merge.
2838 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
2840         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
2841         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
2842         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
2843         (cypress_costs): Set it.
2844         (supersparc_costs): Likewise.
2845         (hypersparc_costs): Likewise.
2846         (leon_cost): Likewise.
2847         (leon3_costs): Likewise.
2848         (sparclet_costs): Likewise.
2849         (ultrasparc_costs): Likewise.
2850         (ultrasparc_costs): Likewise.
2851         (niagara_costs): Likewise.
2852         (niagara2_costs): Likewise.
2853         (niagara3_costs): Likewise.
2854         (niagara4_costs): Likewise.
2855         (niagara7_costs): Likewise.
2856         (m8_costs): Likewise.
2857         (TARGET_CAN_FOLLOW_JUMP): Define.
2858         (pass_work_around_errata::gate): Minor tweak.
2859         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
2860         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
2861         Automaitcally clear MASK_FSMULD mask for V7 processors.
2862         (sparc_can_follow_jump): New static function.
2863         (output_ubranch): Deal with CROSSING_JUMP_P.
2864         (sparc_use_sched_lookahead): Rewrite using switch statement.
2865         (sparc_issue_rate): Reorder.
2866         (sparc_branch_cost): New function.
2868 2018-09-27  Martin Sebor  <msebor@redhat.com>
2870         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
2871         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
2872         (int_fits_type_p): Same.
2874 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
2876         * config/i386/i386.md (FPCR_REG): Remove.
2877         (UNSPEC_FLDCW): Remove.
2878         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
2879         (x86_fldcw_1): Remove insn pattern.
2880         (fnstenv): Do not clobber FPCR_REG.
2881         (fldenv): Ditto.
2882         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
2883         (CALL_USED_REGISTERS): Ditto.
2884         (REG_ALLOC_ORDER): Ditto.
2885         (REG_CLASS_CONTENTS): Ditto.
2886         (HI_REGISTER_NAMES): Ditto.
2887         (ADDITIONAL_REGISTER_NAMES): Use defines instead
2888         of numerical constants.
2889         * config/i386/i386.c (regclass_map): Remove fpsr register.
2890         (dbx_register_map): Ditto.
2891         (dbx64_register_map): Ditto.
2892         (svr4_dbx_register_map): Ditto.
2893         (print_reg): Do not handle FPCR_REG.
2895 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
2897         PR target/87149
2898         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
2899         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
2900         Delete, always treat as true.
2901         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
2902         Ditto.  Simplify remaining code.
2903         * config/powerpcspe/powerpcspe.h: Ditto.
2904         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
2905         Simplify remaining code.
2906         (rs6000_expand_builtin): Ditto.
2907         * config/rs6000/rs6000.h: Ditto.
2908         * configure.ac: Ditto.
2909         * configure: Regenerate.
2911 2018-09-27  Martin Liska  <mliska@suse.cz>
2913         * coverage.c (get_coverage_counts): Revert the formatting
2914         of missing profile opt info.
2916 2018-09-27  Richard Biener  <rguenther@suse.de>
2918         PR debug/37801
2919         PR debug/87440
2920         * dwarf2out.c (set_block_origin_self): Do not mark outermost
2921         block as we do not output that.
2922         (gen_inlined_subroutine_die): Elide the originally outermost
2923         block, matching what we do for concrete instances.
2924         (decls_for_scope): Add parameter specifying whether to recurse
2925         to subblocks.
2927 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
2928             Tom de Vries  <tom@codesourcery.com>
2930         PR 82089
2932         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
2933         STORE_FLAG_VALUE == 1.
2935 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
2937         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
2938         constant definitions.
2939         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
2940         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
2941         ("speculation_barrier"): New expander definition.
2943 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
2945         PR gcov-profile/86957
2946         * common.opt: New warning option -Wmissing-profile.
2947         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
2948         * doc/invoke.texi: Document -Wmissing-profile.
2950 2018-09-26  Jim Wilson  <jimw@sifive.com>
2952         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
2953         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
2955 2018-09-26  Martin Sebor  <msebor@redhat.com>
2957         * tree.c (zerop): Change return type to bool.
2958         (integer_zerop, integer_onep, integer_each_onep): Same.
2959         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
2960         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
2961         (real_onep, real_minus_onep, chain_index): Same.
2962         (print_type_hash_statistics, type_list_equal): Same.
2963         * tree.h (zerop): Same.
2964         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
2965         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
2966         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
2967         (real_onep, real_minus_onep, chain_index): Same.
2968         (print_type_hash_statistics, type_list_equal): Same.
2970 2018-09-26  Jim Wilson  <jimw@sifive.com>
2972         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
2974 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
2976         PR target/87414
2977         * config/i386/i386.c: Include debug.h and dwarf2out.h.
2978         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
2979         call.
2981 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
2983         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
2985 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
2987         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
2988         and FP_SECOND_SSE_REGS.
2989         (REG_CLASS_NAMES): Ditto.
2990         (REG_CLASS_CONTENTS): Ditto.
2991         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
2992         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
2993         (ix86_preferred_output_reload_class): Ditto.
2994         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
2995         clobber constraint to "=&f".
2996         (fix_truncdi_i387): Ditto.
2997         (lrintxfdi2): Ditto.
2998         (fistdi2_<rounding>): Ditto.
2999         (fpremxf4_i387): Change "=u" constraint to "=f".
3000         (fprem1xf4_i387): Ditto.
3001         (sincosxf3): Ditto.
3002         (fptanxf4_i387): Ditto.
3003         (fxtractxf3_i387): Ditto.
3004         (fscalexf4_i387): Ditto.
3005         (atan2xf3): Change "u" constraint to "f".
3006         (fyl2xxf3_i387): Ditto.
3007         (fyl2xp1xf3_i387): Ditto.
3009 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
3011         PR target/87439
3012         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
3013         for removed I387_MASK_PM entity.
3016 2018-09-26  Jeff Law  <law@redhat.com>
3017         Revert
3018         2018-09-26  Alexey Neyman  <stilor@att.net>
3020         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3021         headers are no longer pulled in by <isl/val.h>.
3023 2018-09-26  Richard Biener  <rguenther@suse.de>
3025         PR debug/87443
3026         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
3027         or concrete instance DIE to the tree.  Create abstract origin
3028         attributes also for concrete instances.
3030 2018-09-26  Alexey Neyman  <stilor@att.net>
3032         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
3033         headers are no longer pulled in by <isl/val.h>.
3035 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
3037         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
3038         Use new helper functions.
3039         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
3040         Use new helper functions.
3041         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
3042         aarch_mm_needs_release): New declarations.
3043         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
3044         aarch_mm_needs_release): New.
3046 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
3048         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
3049         (arm32_output_mi_thunk): Deal with long calls.
3051 2018-09-26  Richard Biener  <rguenther@suse.de>
3053         PR debug/87428
3054         PR debug/87362
3055         * tree-inline.c (expand_call_inline): When the location
3056         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
3057         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
3058         the inserted BLOCK to make inlined_function_outer_scope_p
3059         recognize it.
3060         * dwarf2out.c (add_call_src_coords_attributes): Do not add
3061         coords for reserved locations.
3063 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3065         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
3066         (*call_indirect_nonlocal_sysv<mode>): Ditto.
3067         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
3068         (*sibcall_nonlocal_sysv<mode>): Ditto.
3069         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
3070         (<bd>_<mode>): Ditto.
3071         (<bd>tf_<mode>): Ditto.
3073 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3075         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
3076         control string as a list of templates instead of as C code.
3077         (*altivec_movti): Ditto.
3078         * config/rs6000/darwin.md (movdf_low_di): Ditto.
3080 2018-09-25  Jim Wilson  <jimw@sifive.com>
3082         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
3083         when target symbol is weak.
3085 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3087         PR c/87387
3088         * builtins.c (unterminated_array): Simplify.
3089         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
3090         where pointer arithmetic is safe.
3092 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
3094         PR target/86987
3095         * config/rs6000/altivec.md (altivec_vspltb): Use
3096         const_0_to_15_operand instead of u5bit_cint_operand.
3097         (*altivec_vspltb_internal): Ditto.
3098         (altivec_vspltb_direct): Ditto.
3099         (altivec_vsplth): Use const_0_to_7_operand instead of
3100         u5bit_cint_operand.
3101         (*altivec_vsplth_internal): Ditto.
3102         (altivec_vsplth_direct): Ditto.
3103         (altivec_vspltw): Use const_0_to_3_operand instead of
3104         u5bit_cint_operand.
3105         (*altivec_vspltw_internal): Ditto.
3106         (altivec_vspltw_direct): Ditto.
3107         (altivec_vspltsf): Ditto.
3108         (*altivec_vspltsf_internal): Ditto.
3109         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
3110         various splats with the proper size immediate.  Reorder the various
3111         cases by ascending size of immediate, and put all such together.
3113 2018-09-25  Richard Biener  <rguenther@suse.de>
3115         PR debug/83941
3116         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
3117         GC-ification.
3118         (maybe_create_die_with_external_ref): Do not create
3119         DW_TAG_imported_unit here.
3120         (add_abstract_origin_attribute): Handle external BLOCK refs.
3121         (dwarf2out_abstract_function): Simplify LTO case.
3122         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
3123         rather than using maybe_create_die_with_external_ref.
3125 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
3127         PR target/71278
3128         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
3129         (frndintxf2_mask_pm_i387): Ditto.
3130         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
3131         Enable for !flag_trapping_math.
3132         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
3133         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
3134         Change operand 1 predicate to nonimmediate_operand.
3135         (attr "i387_cw"): Remove mask_pm.
3136         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
3137         (enum ix86_entity): Remove I387_MASK_PM.
3138         * config/i386/i386.c (ix86_i387_mode_needed): Do not
3139         handle I387_MASK_PM.
3140         (ix86_mode_needed): Ditto.
3141         (ix86_mode_after): Ditto.
3142         (ix86_mode_entry): Ditto.
3143         (ix86_mode_exit): Ditto.
3144         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
3146 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
3148         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
3149         to_update_switch_stmts to vNULL instead of calling create on them
3150         immediately.
3152 2018-09-25  Richard Biener  <rguenther@suse.de>
3154         PR tree-optimization/87402
3155         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
3156         (visit_phi): Re-instantiate handling of supposed to be VARYING
3157         but non-VARYING backedge value.
3159 2018-09-25  Richard Biener  <rguenther@suse.de>
3161         PR debug/83941
3162         * dwarf2out.c (struct sym_off_pair): New.
3163         (external_die_map): New global.
3164         (lookup_decl_die): When in LTO create DIEs lazily from the
3165         external_die_map.
3166         (lookup_block_die): New function, create DIEs lazily in LTO.
3167         (equate_block_to_die): New function.
3168         (dwarf2out_die_ref_for_decl): During WPA get the association
3169         from the external DIE map.
3170         (dwarf2out_register_external_die): Record mapping into the
3171         external DIE map.
3172         (maybe_create_die_with_external_ref): New function split out from
3173         DIE generation part of old dwarf2out_register_external_die.
3174         (add_abstract_origin_attribute): Do not return the DIE.  When
3175         in LTO reference externals directly.
3176         (dwarf2out_abstract_function): When in LTO ignore calls for
3177         decls with external DIEs (already present abstract instances).
3178         (gen_call_site_die): Adjust.
3179         (add_high_low_attributes): Likewise.
3180         (gen_lexical_block_die): Likewise.
3181         (gen_inlined_subroutine_die): Likewie.
3182         (gen_block_die): Likewise.
3183         (dwarf2out_inline_entry): Likewise.
3184         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
3185         DIEs.
3187 2018-09-25  Martin Liska  <mliska@suse.cz>
3189         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
3190         integers and not by a float value.
3192 2018-09-25  Martin Liska  <mliska@suse.cz>
3194         PR fortran/87394
3195         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
3196         instead of NULL.
3197         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
3198         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
3199         * godump.c (go_format_type): Remove extra parenthesis.
3201 2018-09-25  Martin Liska  <mliska@suse.cz>
3203         * alias.c (set_dest_equal_p): Remove unused function.
3204         * config/i386/i386.c (def_builtin_pure2): Likewise.
3205         * diagnostic-show-locus.c (class layout): Remove
3206         unused field.
3207         (layout::layout): Likewise here.
3208         * dump-context.h (class temp_dump_context): Likewise.
3209         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
3210         (add_AT_loclistsptr): Likewise.
3211         (add_AT_offset): Likewise.
3212         (get_AT_hi_pc): Likewise.
3213         (is_comdat_die): Likewise.
3214         (type_is_enum): Likewise.
3215         (ceiling): Likewise.
3216         (add_AT_vms_delta): Likewise.
3217         (is_class_die): Likewise.
3218         * edit-context.c (class line_event): Remove unused field.
3219         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
3220         unused function.
3221         * ipa-cp.c (ipa_get_vr_lat): Likewise.
3222         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
3223         (ok_for_base_p_nonstrict): Likewise.
3224         * tree-chrec.c (is_not_constant_evolution): Likewise.
3225         (chrec_fold_poly_cst): Likewise.
3226         * tree-if-conv.c (has_pred_critical_p): Likewise.
3227         * tree-ssa-coalesce.c (print_exprs): Likewise.
3228         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
3229         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
3230         * tree-vrp.c (value_ranges_intersect_p): Likewise.
3231         (value_range_nonnegative_p): Likewise.
3233 2018-09-25  Martin Liska  <mliska@suse.cz>
3235         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
3236         Do not handle "GNU Pascal".
3237         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
3238         Likewise.
3239         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
3240         from documentation. Likewise.
3241         * dbxout.c (dbxout_range_type): Likewise.
3242         * doc/cpp.texi: Likewise.
3243         * doc/extend.texi: Likewise.
3244         * doc/frontends.texi: Likewise.
3245         * doc/invoke.texi: Remove Pascal entry.
3246         * tree.def (CLEANUP_POINT_EXPR): Likewise.
3247         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
3249 2018-09-25  Martin Liska  <mliska@suse.cz>
3251         PR middle-end/86078
3252         * doc/invoke.texi: Document all parameters and remove default
3253         of the parameters.
3255 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3257         PR bootstrap/87417
3258         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
3259         contains HOST_WIDE_INTs when computing its size.
3261 2018-09-24  Jim Wilson  <jimw@sifive.com>
3263         PR target/87391
3264         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
3265         not TARGET_RVE.
3266         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
3268 2018-09-24  Andrew Pinski  <apinski@marvell.com>
3270         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
3271         access prev before checking it for NULLness in the
3272         AARCH64_FUSE_CMP_BRANCH case.
3274 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
3276         PR target/82699
3277         * config/i386/i386.c (rest_of_insert_endbranch): Set
3278         endbr_queued_at_entrance to true and don't insert ENDBR if
3279         x86_function_profiler will be called.
3280         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
3281         is true.
3282         * config/i386/i386.h (machine_function): Add
3283         endbr_queued_at_entrance.
3285 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3287         * genattrtab.c (mk_attr_alt): Use alternative_mask.
3288         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
3289         types.
3290         (check_attr_test): Use alternative_mask.
3291         (get_attr_value): Likewise.
3292         (compute_alternative_mask): Use alternative_mask and XWINT.
3293         (make_alternative_compare): Use alternative_mask.
3294         (attr_alt_subset_p): Use XWINT.
3295         (attr_alt_subset_of_compl_p): Likewise.
3296         (attr_alt_intersection): Use alternative_mask and XWINT.
3297         (attr_alt_union): Likewise.
3298         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
3299         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
3300         (simplify_test_exp): Use alternative_mask and XWINT.
3301         (write_test_expr): Use alternative_mask and XWINT, adjust bit
3302         number calculation to support 64 bits.  Generate code that
3303         checks 64-bit masks.
3304         (main): Use alternative_mask.
3305         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
3307 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3309         PR target/80080
3310         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
3311         RETURN+USE when returning via %r14.
3313 2018-09-24  Martin Liska  <mliska@suse.cz>
3315         * gcov.c (output_lines): Print colorization legend
3316         for both flag_use_colors and flag_use_hotness_colors.
3317         Reword the help.
3319 2018-09-24  Martin Liska  <mliska@suse.cz>
3321         * coverage.c (get_coverage_counts): Use warning_at
3322         with current_function_decl location. Use %qD in warning
3323         message.
3325 2018-09-24  Martin Liska  <mliska@suse.cz>
3327         * memory-block.h (memory_block_pool::release): Annotate with
3328         valgrind that the memory is not accessible.
3330 2018-09-24  Martin Liska  <mliska@suse.cz>
3332         PR sanitizer/85774
3333         * asan.c: Make asan_handled_variables extern.
3334         * asan.h: Likewise.
3335         * cfgexpand.c (expand_stack_vars): Make sure
3336         a representative is unpoison if another
3337         variable in the partition is handled by
3338         use-after-scope sanitization.
3340 2018-09-24  Richard Biener  <rguenther@suse.de>
3342         PR tree-optimization/63155
3343         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
3344         the worklist when the edge of the respective argument isn't
3345         executable.
3347 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
3349         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
3350         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
3351         (MASK_CLASS_P): Update for rename.
3352         (MAYBE_MASK_CLASS_P): Ditto.
3353         (REG_CLASS_NAMES): Update.
3354         (REG_CLASS_CONTENT): Update.
3355         * config/i386/i386.c (regclass_map): Update for MASK_REG
3356         and ALL_MASK_REGS rename.
3357         * config/i386/constraints.md (Yk): Update for rename.
3358         (k): Ditto.
3360 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
3362         * config/i386/i386.h (enum reg_class): Remove
3363         EVEX_SSE_REGS and MOD4_SSE_REGS.
3364         (REG_CLASS_NAMES): Update.
3365         (REG_CLASS_CONTENT): Update.
3366         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
3367         registers as ALL_SSE_REGS.
3368         (ix86_additional_allocno_class_p): Remove.
3369         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
3370         (ix86_register_priority): Lower priority of EVEX SSE registers.
3371         Use IN_RANGE macro where appropriate.
3372         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
3373         AVX-5124VNNIW checks.
3374         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
3375         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
3376         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
3377         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
3378         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
3379         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
3380         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
3381         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
3382         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
3383         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
3384         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
3385         * config/i386/constraints.md (Yh): Remove.
3387 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
3389         * config/i386/i386.c (regclass_map): Declare integer REX registers
3390         as GENERAL_REGS.
3392 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
3394         * doc/service.texi (Service): Switch the fsf.org link to https.
3396 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
3398         PR target/86798
3399         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3400         Define to speculation_safe_value_not_needed.
3402 2018-09-21  Florian Weimer  <fweimer@redhat.com>
3404         PR middle-end/81035
3405         * doc/extend.texi (Common Function Attributes): Mention that
3406         noreturn suppresses tail call optimization.
3408 2018-09-21  Jeff Law  <law@redhat.com>
3410         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
3411         vr_values::cleanup_edges_and_switches.
3412         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
3413         vr_values class.
3414         (identify_jump_threads): Remove EDGE_IGNORE handling.
3415         (execute_vrp): Move handling of to_remove_edges and
3416         to_update_switch_stmts into vr_values class member functions.
3417         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
3418         (to_update_switch_stmts): Likewise.
3419         * vr-values.c: Include cfghooks.h.
3420         (vr_values::vr_values): Initialize to_remove_edges and
3421         to_update_switch_stmts.
3422         (vr_values::~vr_values): Verify to_remove_edges and
3423         to_update_switch_stmts are empty.
3424         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
3425         (vr_values::cleanup_edges_and_switches): New member function.
3426         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
3427         function.  Add new data members.
3429 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
3431         PR tree-optimization/87309
3432         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
3433         calls with pflags and alt_flags.
3434         (selftest::test_capture_of_dump_calls): Add test of interaction of
3435         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
3437 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3439         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
3441 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3443         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
3445 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3447         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
3448         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
3449         (VXWORKS_LIBS_RTP): Minor reordering.
3451 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3453         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
3454         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
3456 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3458         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
3459         (PTRDIFF_TYPE): Likewise.
3461 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3463         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
3464         triplet, similar to support for VxWorks7.
3465         * config/vxworks-dummy.h: Provide a default definition
3466         of TARGET_VXWORKS64 to 0.
3468 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3470         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
3471         * config/vxworks-dummy.h: here.
3473 2018-09-21  Olivier Hainque  <hainque@adacore.com>
3475         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
3477 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
3478             Bo Zhou  <zbo.zhou@hisilicon.com>
3480         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
3481         * config/aarch64/aarch64-tune.md: Regenerated.
3482         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
3483         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
3484         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
3486 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
3487             Julian Brown  <julian@codesourcery.com>
3489         * builtins.c (get_builtin_sync_mem): Handle address spaces.
3491 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
3493         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
3494         if the call takes a static chain.
3496 2018-09-21  Martin Liska  <mliska@suse.cz>
3498         * auto-profile.c (autofdo_source_profile::read): Do not
3499         set sum_all.
3500         (read_profile): Do not add working sets.
3501         (read_autofdo_file): Remove sum_all.
3502         (afdo_callsite_hot_enough_for_early_inline): Remove const
3503         qualifier.
3504         * coverage.c (struct counts_entry): Remove gcov_summary.
3505         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
3506         do not support GCOV_TAG_PROGRAM_SUMMARY.
3507         (get_coverage_counts): Remove summary and expected
3508         arguments.
3509         * coverage.h (get_coverage_counts): Likewise.
3510         * doc/gcov-dump.texi: Remove -w option.
3511         * gcov-dump.c (dump_working_sets): Remove.
3512         (main): Do not support '-w' option.
3513         (print_usage): Likewise.
3514         (tag_summary): Likewise.
3515         * gcov-io.c (gcov_write_summary): Do not dump
3516         histogram.
3517         (gcov_read_summary): Likewise.
3518         (gcov_histo_index): Remove.
3519         (gcov_histogram_merge): Likewise.
3520         (compute_working_sets): Likewise.
3521         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
3522         it not obsolete.
3523         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
3524         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
3525         (GCOV_HISTOGRAM_SIZE): Remove.
3526         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
3527         (struct gcov_summary): Simplify rapidly just
3528         to runs and sum_max fields.
3529         (gcov_histo_index): Remove.
3530         (NUM_GCOV_WORKING_SETS): Likewise.
3531         (compute_working_sets): Likewise.
3532         * gcov-tool.c (print_overlap_usage_message): Remove
3533         trailing empty line.
3534         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
3535         (output_lines): Remove program related line.
3536         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
3537         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
3538         histogram.
3539         (input_profile_summary): Do not read it.
3540         (merge_profile_summaries): And do not merge it.
3541         (input_symtab): Do not call removed function.
3542         * modulo-sched.c (sms_schedule): Do not print sum_max.
3543         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
3544         removed when histogram method was invented.
3545         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
3546         mode.
3547         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
3548         GCOV coding style.
3549         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
3550         and dump selected value.
3551         * profile.c (add_working_set): Remove.
3552         (get_working_sets): Likewise.
3553         (find_working_set): Likewise.
3554         (get_exec_counts): Do not work with working sets.
3555         (read_profile_edge_counts): Do not inform as sum_max is removed.
3556         (compute_branch_probabilities): Likewise.
3557         (compute_value_histograms): Remove argument for call of
3558         get_coverage_counts.
3559         * profile.h: Do not make gcov_summary const.
3561 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
3563         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
3565 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
3567         PR tree-optimization/86990
3568         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
3569         Check that the entire merged store group is made of constants only for
3570         overlapping stores.
3572 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
3574         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
3575         (VTABLE_VERIFICATION_SPEC): Likewise.
3576         (SANITIZER_EARLY_SPEC): Likewise.
3577         (SANITIZER_SPEC): Likewise.
3578         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
3579         * doc/invoke.texi (Link Options): Document -r.
3581 2018-09-20  Richard Biener <rguenther@suse.de>
3583         PR middle-end/87054
3584         * gimplify.c (gimplify_expr): Retain alignment of
3585         addressable lvalue in dereference.
3587 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
3589         PR bootstrap/87013
3590         * configure.ac: Check for .loc is_stmt support.
3591         * configure, config.in: Rebuilt.
3592         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
3593         if not supported.
3595 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
3597         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
3598         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
3599         -misel=no.
3601 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
3603         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3604         VECTOR_OTHER.
3605         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
3606         case VECTOR_OTHER.
3608 2018-09-20  Marek Polacek  <polacek@redhat.com>
3610         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
3612 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
3614         PR tree-optimization/87288
3615         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
3616         into account when determining PEELING_FOR_NITERS.
3618 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
3620         PR tree-optimization/86877
3621         * tree-vect-loop.c (vect_analyze_loop_2): Call
3622         vect_verify_datarefs_alignment.
3624 2018-09-19  Marek Polacek  <polacek@redhat.com>
3626         * doc/invoke.texi: Document -Wclass-conversion.
3628 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
3630         * config/pa/pa.c (pa_adjust_priority): Delete.
3631         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
3633         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
3634         (atomic_storehi): Likewise.
3635         (atomic_storesi): Likewise.
3636         (atomic_loaddi): Restore compare and swap exchange loop code.
3638 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
3640         PR rtl-optimization/86902
3641         * combine.c (try_combine): When changing the CC mode used, don't change
3642         an unrelated mode in other_insn to that new CC mode.
3644 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
3646         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
3647         with %T in place of calls to dump_generic_expr.
3648         (prune_runtime_alias_test_list): Likewise.
3649         (create_runtime_alias_checks): Likewise.
3650         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
3651         (vect_analyze_data_ref_dependence): Likewise.
3652         (vect_slp_analyze_data_ref_dependence): Likewise.
3653         (vect_record_base_alignment): Likewise.  Use %G in place of call
3654         to dump_gimple_stmt.
3655         (vect_compute_data_ref_alignment): Likewise.
3656         (verify_data_ref_alignment): Likewise.
3657         (vect_find_same_alignment_drs): Likewise.
3658         (vect_analyze_group_access_1): Likewise.
3659         (vect_analyze_data_ref_accesses): Likewise.
3660         (dependence_distance_ge_vf): Likewise.
3661         (dump_lower_bound): Likewise.
3662         (vect_prune_runtime_alias_test_list): Likewise.
3663         (vect_find_stmt_data_reference): Likewise.
3664         (vect_analyze_data_refs): Likewise.
3665         (vect_create_addr_base_for_vector_ref): Likewise.
3666         (vect_create_data_ref_ptr): Likewise.
3667         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
3668         (vect_can_advance_ivs_p): Likewise.
3669         (vect_update_ivs_after_vectorizer): Likewise.
3670         (vect_gen_prolog_loop_niters): Likewise.
3671         (vect_prepare_for_masked_peels): Likewise.
3672         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
3673         (vect_determine_vectorization_factor): Likewise.
3674         (vect_is_simple_iv_evolution): Likewise.
3675         (vect_analyze_scalar_cycles_1): Likewise.
3676         (vect_analyze_loop_operations): Likewise.
3677         (report_vect_op): Likewise.
3678         (vect_is_slp_reduction): Likewise.
3679         (check_reduction_path): Likewise.
3680         (vect_is_simple_reduction): Likewise.
3681         (vect_create_epilog_for_reduction): Likewise.
3682         (vect_finalize_reduction:): Likewise.
3683         (vectorizable_induction): Likewise.
3684         (vect_transform_loop_stmt): Likewise.
3685         (vect_transform_loop): Likewise.
3686         (optimize_mask_stores): Likewise.
3687         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
3688         (vect_split_statement): Likewise.
3689         (vect_recog_over_widening_pattern): Likewise.
3690         (vect_recog_average_pattern): Likewise.
3691         (vect_determine_min_output_precision_1): Likewise.
3692         (vect_determine_precisions_from_range): Likewise.
3693         (vect_determine_precisions_from_users): Likewise.
3694         (vect_mark_pattern_stmts): Likewise.
3695         (vect_pattern_recog_1): Likewise.
3696         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3697         (vect_record_max_nunits): Likewise.
3698         (vect_build_slp_tree_1): Likewise.
3699         (vect_build_slp_tree_2): Likewise.
3700         (vect_print_slp_tree): Likewise.
3701         (vect_analyze_slp_instance): Likewise.
3702         (vect_detect_hybrid_slp_stmts): Likewise.
3703         (vect_detect_hybrid_slp_1): Likewise.
3704         (vect_slp_analyze_operations): Likewise.
3705         (vect_slp_analyze_bb_1): Likewise.
3706         (vect_transform_slp_perm_load): Likewise.
3707         (vect_schedule_slp_instance): Likewise.
3708         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
3709         (vect_mark_stmts_to_be_vectorized): Likewise.
3710         (vect_init_vector_1): Likewise.
3711         (vect_get_vec_def_for_operand): Likewise.
3712         (vect_finish_stmt_generation_1): Likewise.
3713         (vect_check_load_store_mask): Likewise.
3714         (vectorizable_call): Likewise.
3715         (vectorizable_conversion): Likewise.
3716         (vectorizable_operation): Likewise.
3717         (vectorizable_load): Likewise.
3718         (vect_analyze_stmt): Likewise.
3719         (vect_is_simple_use): Likewise.
3720         (vect_get_vector_types_for_stmt): Likewise.
3721         (vect_get_mask_type_for_stmt): Likewise.
3722         * tree-vectorizer.c (increase_alignment): Likewise.
3724 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
3726         * doc/rtl.texi: Adjust vec_select description.
3727         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
3728         non-constant selectors.
3730 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
3732         * config/aarch64/aarch64-protos.h
3733         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
3734         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
3735         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
3736         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
3737         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
3738         (AARCH64_FL_PROFILE): Move index so flags are ordered.
3739         (AARCH64_ISA_RCPC8_4): New flag.
3740         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
3741         to aarch64_offset_9bit_signed_unscaled_p.
3742         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
3743         and use stlur.
3744         * config/aarch64/constraints.md (Ust): New constraint.
3745         * config/aarch64/predicates.md.
3746         (aarch64_9bit_offset_memory_operand): New predicate.
3747         (aarch64_rcpc_memory_operand): New predicate.
3749 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
3751         PR rtl-optimization/87361
3752         * rtlanal.c (nonzero_bits1): Revert accidental change.
3754 2018-09-19  Richard Biener  <rguenther@suse.de>
3756         PR tree-optimization/87349
3757         PR tree-optimization/87342
3758         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
3760 2018-09-18  Marek Polacek  <polacek@redhat.com>
3762         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
3763         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
3765 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
3767         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
3769 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
3771         PR rtl-optimization/86882
3772         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
3774 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
3776         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
3777         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
3779 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
3781         PR other/87353
3782         * doc/invoke.texi (Link Options): Fix formatting and grammar.
3784 2018-09-18  Richard Biener  <rguenther@suse.de>
3786         PR middle-end/63155
3787         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
3788         (compute_samebase_partition_bases): Likewise.
3789         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
3790         (gimple_can_coalesce_p): Simplify.
3792 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
3794         Handle a library implementation of ffs calling __builtin_ffs.
3795         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
3796         (mmix_init_libfuncs): New function: make __builtin_ffs expand
3797         to __ffsdi2.
3799 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
3801         * diagnostic-show-locus.c (class layout_range): Add field
3802         "m_original_idx".
3803         (layout_range::layout_range): Add "original_idx" param and use it
3804         to initialize new field.
3805         (make_range): Use 0 for original_idx.
3806         (layout::layout): Pass in index to calls to
3807         maybe_add_location_range.
3808         (layout::maybe_add_location_range): Add param "original_idx" and
3809         pass it on to layout_range.
3810         (layout::print_any_labels): Pass on range->m_original_idx to
3811         get_text call.
3812         (gcc_rich_location::add_location_if_nearby): Use 0 for
3813         original_idx.
3814         * gcc-rich-location.h (text_range_label::get_text): Update for new
3815         param.
3816         (range_label_for_type_mismatch::get_text): Likewise.
3818 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
3820         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
3822 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
3824         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
3825         format_string_diagnostic_t.
3826         (fmtwarn_n): Likewise.
3827         * substring-locations.c
3828         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
3829         (format_warning_n_va): Convert to...
3830         (format_string_diagnostic_t::emit_warning_n_va): ...this.
3831         (format_warning_va): Convert to...
3832         (format_string_diagnostic_t::emit_warning_va): ...this.
3833         (format_warning_at_substring): Convert to...
3834         (format_string_diagnostic_t::emit_warning): ...this.
3835         (format_warning_at_substring_n): Convert to...
3836         (format_string_diagnostic_t::emit_warning_n): ...this.
3837         * substring-locations.h (class format_string_diagnostic_t): New
3838         class.
3839         (format_warning_va): Convert to
3840         format_string_diagnostic_t::emit_warning_va.
3841         (format_warning_n_va): Convert to
3842         format_string_diagnostic_t::emit_warning_n_va.
3843         (format_warning_at_substring): Convert to
3844         format_string_diagnostic_t::emit_warning.
3845         (format_warning_at_substring_n): Convert to
3846         format_string_diagnostic_t::emit_warning_n.
3848 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
3849             Bernd Schmidt <bernds_cb1@t-online.de>
3851         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
3852         SImode args.
3854 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
3856         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
3857         operand 0 predicate to nonimmediate operand.
3858         (rint<mode>2_frndint): Remove insn pattern.
3859         (rint<mode>2): Change operand 1 predicate to general_operand.
3860         Extend operand 1 to XFmode and generate rintxf2 insn.
3861         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
3862         Do not use X87MODEF mode macro.
3863         (frndintxf2_<rounding>_i387): Rename from
3864         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
3865         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
3866         to XFmode and generate significandxf3 insn.
3868 2018-09-17  Richard Biener  <rguenther@suse.de>
3870         PR tree-optimization/87328
3871         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
3872         visiting unexecutable backedges when not iterating.
3873         (do_rpo_vn): Mark all edges not executable even when not
3874         iterating.
3876 2018-09-17  Martin Jambor  <mjambor@suse.cz>
3878         PR c/63886
3879         * doc/invoke.texi (Warning Options): Likewise.
3881 2018-09-17  Richard Biener  <rguenther@suse.de>
3883         PR tree-optimization/87301
3884         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
3885         clean EH info from leftover copy assignments.
3887 2018-09-17  Martin Liska  <mliska@suse.cz>
3889         PR gcov-profile/85871
3890         * gcov.c (output_intermediate_file): Fix out of bounds
3891         access.
3893 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
3895         * config/arc/arc.c: Object attributes for core4 not reflected
3896         correctly.
3897         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
3898         core3).
3900 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
3902         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
3904 2018-09-17  Martin Liska  <mliska@suse.cz>
3906         * doc/gcov.texi: Document new option --use-hotness-colors.
3907         * gcov.c (struct source_info): Declare new field.
3908         (source_info::source_info): Set default for maximum_count.
3909         (print_usage): Add new -q option.
3910         (process_args): Process it.
3911         (accumulate_line_info): Save src->maximum_count.
3912         (output_line_beginning): Make color line number if
3913         flag_use_hotness_colors is set.
3914         (output_line_details): Pass default argument value.
3915         (output_lines): Pass src->maximum_count.
3917 2018-09-17  Martin Liska  <mliska@suse.cz>
3919         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
3920         Use processor_names table.
3921         * config/i386/i386.c (ix86_default_align): Use
3922         processor_cost_table for alignment values.
3923         (ix86_option_override_internal): Use processor_names.
3924         (ix86_function_specific_print): Likewise.
3925         * config/i386/i386.h (struct processor_costs):
3926         Add alignment values.
3927         (struct ptt): Remove and replace with const char *.
3928         * config/i386/x86-tune-costs.h (struct processor_costs):
3929         Declare default alignments for all costs.
3931 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
3933         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
3934         symbolics or VR_VARYING ranges for ABS_EXPR.
3935         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
3936         when range will wrap.
3938 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
3940         PR middle-end/86864
3941         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
3942         before and after a JUMP_TABLE_DATA.
3944 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
3946         PR middle-end/87188
3947         * dojump.c (do_compare_and_jump): Canonicalize function pointers
3948         when one operand is a function pointer.  Use POINTER_TYPE_P and
3949         FUNC_OR_METHOD_TYPE_P.
3950         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
3951         * fold-const.c (build_range_check): Likewise.
3952         * match.pd (simple_comparison): Likewise.
3954 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
3956         PR c/82967
3957         * spellcheck.c (get_edit_distance_cutoff): New function.
3958         (selftest::test_edit_distance_unit_test_oneway): Rename to...
3959         (selftest::test_get_edit_distance_one_way): ...this.
3960         (selftest::test_get_edit_distance_unit): Rename to...
3961         (selftest::test_get_edit_distance_both_ways): ...this.
3962         (selftest::test_edit_distances): Move tests to this new function,
3963         and test some more pairs of strings.  Update for above renaming.
3964         (selftest::get_old_cutoff): New function.
3965         (selftest::test_get_edit_distance_cutoff): New function.
3966         (selftest::assert_suggested_for): New function.
3967         (ASSERT_SUGGESTED_FOR): New macro.
3968         (selftest::assert_not_suggested_for): New function.
3969         (ASSERT_NOT_SUGGESTED_FOR): New macro.
3970         (selftest::test_suggestions): New function.
3971         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
3972         tests to selftest::test_edit_distances and call it.  Add calls to
3973         selftest::test_get_edit_distance_cutoff and
3974         selftest::test_suggestions.
3975         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
3976         (best_match::consider): Replace hard-coded cutoff calculation with
3977         a call to...
3978         (best_match::get_cutoff): New declaration.
3979         (best_match::get_best_meaningful_candidate): Likewise.
3981 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3983         * builtins.c (fold_builtin_strlen): Remove TODO comment.
3985 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3987         revert:
3988         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3990         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
3991         terminated string literal.
3993 2018-09-14  Martin Sebor  <msebor@redhat.com>
3995         * builtins.c (unterminated_array): Handle ARRAY_REF.
3996         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
3997         * builtins.h (unterminated_array): Declare extern.
3998         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
3999         arrays.
4000         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
4001         calls.
4003 2018-09-14  Martin Sebor  <msebor@redhat.com>
4004             Jeff Law  <law@redhat.com>
4006         * builtins.c (unterminated_array): New.
4007         (expand_builtin_strcpy): Adjust.
4008         (expand_builtin_strcpy_args): Detect unterminated arrays.
4009         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
4010         unterminated arrays.
4011         * gimple-fold.h (get_maxval_strlen): Add argument.
4012         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
4014         * gimple-fold.c (get_range_strlen): Add argument.
4015         (get_maxval_strlen): Adjust.
4016         * gimple-fold.h (get_range_strlen): Add argument.
4018 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
4020         * config/i386/movdirintrin.h: Fix copyright year.
4022 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
4024         * reg-stack.c: Include regs.h.
4025         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
4026         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
4027         FIRST_STACK_REG, not DFmode.
4028         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
4029         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
4030         (change stack): Default register mode to the reg_raw_mode of
4031         FIRST_STACK_REG, not DFmode.
4032         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
4033         (*swapxf): Rename from swapxf.
4035 2018-09-14  Carl Love  <cel@us.ibm.com>
4037         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
4038         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
4040 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
4042         PR target/87224
4043         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
4044         alternatives.
4046 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
4048         PR target/85628
4049         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
4051 2018-09-14  Jason Merrill  <jason@redhat.com>
4053         Fix --enable-gather-detailed-mem-stats.
4054         * hash-table.c (hash_table_usage): Change from variable to function.
4055         * hash-table.h: Adjust.
4056         * Makefile.in: Add missing dependencies on hash-table.h.
4058 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4060         PR tree-optimization/87259
4061         PR lto/87283
4062         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
4063         execute_cse_reciprocals_1 has tried transforming.
4065 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
4067         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
4068         VR_VARYING for PLUS/MINUS_EXPR.
4070 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
4072         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
4073         formatting.
4075 2018-09-14  Richard Biener  <rguenther@suse.de>
4077         PR middle-end/63155
4078         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
4079         bits for the merged partition.
4081 2018-09-13  Martin Sebor  <msebor@redhat.com>
4082             Bernd Edlinger  <bernd.edlinger@hotmail.de>
4084         * builtins.h (c_srlen): Add argument.
4085         * builtins.c (warn_string_no_nul): New function.
4086         (c_strlen): Add argument and use it.  Update recursive calls.
4087         Pass DECL argument to string_constant to get info on non
4088         terminated strings.  Update *NONSTR as needed.
4089         (fold_builtin_strlen): Add argument to calls to c_strlen.
4090         Warn for unterminated arrays.
4091         (warn_string_no_null): Add prototype.
4092         * expr.c (string_constant): Update arguments.  Update recursive
4093         calls appropriately.  Detect missing NUL terminator and outermost
4094         declaration its missing in.
4095         Improve checks for arrays with nonzero lower bound or elements
4096         that are not a single byte.  Simplify offset computation.
4097         Simplify checks for non-NUL terminated strings.
4098         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
4099         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
4101 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4103         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
4104         correctly.
4105         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
4106         argument.  Fix range checks.
4107         * fold-const.h (c_getstr): Adjust protoype.
4108         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
4109         string is constant but contains no NUL byte.
4111         * expr.c (string_constant): Adjust function comment.
4112         Remove bogus check for zero termination.
4114         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
4116         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
4117         (get_constant_size): Don't make STRING_CSTs larger than they are.
4118         (check_string_literal): New check function for STRING_CSTs.
4119         (output_constant): Use it.
4121 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
4123         PR target/86812
4124         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
4126 2018-09-13  Richard Biener  <rguenther@suse.de>
4128         PR tree-optimization/87263
4129         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
4130         (struct unwind_state): Add max_rpo field.
4131         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
4132         Compute max_rpo, the max RPO number a block can be backwards reached
4133         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
4134         separating it from the iterating mode.
4136 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
4138         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
4139         (rfs_decision): New scheduling decision.
4141 2018-09-13  Richard Biener  <rguenther@suse.de>
4143         PR bootstrap/87134
4144         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
4145         (vn_nary_op_insert_pieces_predicated): Do not write useless
4146         valid_dominated_by_p entry outside of the allocated storage.
4148 2018-09-13  Omar Sandoval  <osandov@osandov.com>
4149             Tom de Vries  <tdevries@suse.de>
4151         PR debug/86985
4152         * dwarf2out.c (is_c): New function.
4153         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
4155 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
4157         PR target/85628
4158         * config/aarch64/aarch64.md (*aarch64_bfxil):
4159         Define.
4160         * config/aarch64/constraints.md (Ulc): Define.
4161         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
4162         Define.
4163         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
4164         New function.
4166 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
4168         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
4169         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
4170         aarch64_layout_frame call.
4171         (aarch64_expand_epilogue): Likewise.
4172         (aarch64_initial_elimination_offset): Likewise.
4173         (aarch64_get_separate_components): Likewise.
4174         (aarch64_use_return_insn_p): Likewise.
4175         (aarch64_layout_frame): Remove unneeded check.
4177 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4179         * configure.ac: Only append
4180         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
4181         gcc_config_arguments if it was never reconfigured or last reconfigure
4182         was with different arguments.
4183         * configure: Regenerated.
4185 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4186             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4188         PR middle-end/87290
4189         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
4190         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
4192 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
4194         PR tree-optimization/87287
4195         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
4196         X % C == 0 to X % (unsigned) C == 0 optimization to ...
4197         * match.pd (X % C == 0): ... here.  New optimization.
4199 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
4201         PR middle-end/82853
4202         * expr.h (maybe_optimize_mod_cmp): Declare.
4203         * expr.c (mod_inv): New function.
4204         (maybe_optimize_mod_cmp): New function.
4205         (do_store_flag): Use it.
4206         * cfgexpand.c (expand_gimple_cond): Likewise.
4208 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
4209             Julian Brown  <julian@codesourcery.com>
4211         PR middle-end/86336
4212         * gimplify.c (gimplify_scan_omp_clauses): Set
4213         target_firstprivatize_array_bases in OpenACC parallel and kernels
4214         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
4215         OpenACC data regions.
4217 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
4219         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
4220         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
4222 2018-09-12  Richard Biener  <rguenther@suse.de>
4224         PR tree-optimization/87280
4225         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
4226         edge but unreachable target.
4227         (do_rpo_vn): For conservatively handling a PHI only mark
4228         the backedge executable but not the block reachable.
4230 2018-09-12  Richard Biener  <rguenther@suse.de>
4232         PR tree-optimization/87266
4233         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
4234         visited blocks.
4236 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
4238         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
4239         constants.
4240         ("trunc<BFP:mode><DFP_ALL:mode>2")
4241         ("trunc<DFP_ALL:mode><BFP:mode>2")
4242         ("extend<BFP:mode><DFP_ALL:mode>2")
4243         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
4244         according to the target operand type.
4246 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
4247             Andreas Krebbel  <krebbel@linux.ibm.com>
4249         PR tree-optimization/86844
4250         * gimple-ssa-store-merging.c
4251         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
4252         there are any overlapping stores in between them, make sure they are
4253         also coalesced or we give up completely.
4255 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
4257         PR middle-end/87248
4258         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
4259         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
4261 2018-09-12  Tom de Vries  <tdevries@suse.de>
4263         * common.opt (gdescribe-dies): Add option.
4264         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
4265         attribute for artifical and nameless decls.
4266         (dwarf2out_register_external_die): Add description attribute to
4267         external reference die.
4268         (add_desc_attribute): New functions.
4269         (gen_subprogram_die): Add description attribute to
4270         DW_TAG_call_site_parameter.
4271         * tree-pretty-print.c (print_generic_expr_to_str): New function.
4272         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
4273         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
4274         -gno-describe-dies.
4275         (@item -gdescribe-dies): Add.
4277 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
4279         * tree-vrp.c (vrp_shift_undefined_p): Remove.
4280         (extract_range_from_binary_expr_1: Call
4281         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
4282         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
4283         depend on sign.
4285 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
4287         * gimple-ssa-warn-alloca.c
4288         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
4289         field for ALLOCA_BOUND_*_LARGE.
4291 2018-09-11  Nathan Sidwell  <nathan@acm.org>
4293         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
4295 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
4297         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
4298         for clobbers.  Remove obsolete comment.
4300 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
4302         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
4303         mpxchk, mpxld and mpxst types.
4304         (define_attr length_immediate): Remove all processing of mpx types.
4305         (define_attr prefix_0f): Ditto.
4306         (define_attr memory): Ditto.
4308 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
4310         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
4311         (log<mode>2): Change operand 1 predicate to general_operand.
4312         Extend operand 1 to XFmode and generate logxf3 insn.
4313         (log10<mode>2): Change operand 1 predicate to general_operand.
4314         Extend operand 1 to XFmode and generate log10xf3 insn.
4315         (log2<mode>2): Change operand 1 predicate to general_operand.
4316         Extend operand 1 to XFmode and generate log2xf3 insn.
4317         (fyl2xp1_extend<mode>xf3_i387): Remove.
4318         (log1p<mode>2): Change operand 1 predicate to general_operand.
4319         Extend operand 1 to XFmode and generate log1pxf3 insn.
4320         (fxtract_extend<mode>xf3_i387): Remove.
4321         (logb<mode>2): Change operand 1 predicate to general_operand.
4322         Extend operand 1 to XFmode and generate logbxf3 insn.
4323         (ilogb<mode>2): Change operand 1 predicate to general_operand.
4324         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
4325         (significand<mode>2): Change operand 1 predicate to general_operand.
4326         Extend operand 1 to XFmode and generate significandxf3 insn.
4328 2018-09-11  Nathan Sidwell  <nathan@acm.org>
4330         * gcc.c (perror_with_name, pfatal_with_name): Delete.
4331         (load_specs): Use fatal_error.
4332         (DELETE_IF_ORDINARY, process_command): Use error.
4333         (execute, run_attempt): Use fatal_error.
4335 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
4337         * diagnostic-core.h (sorry_at): New prototype.
4338         * diagnostic.c (sorry_at): New function.
4340 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
4342         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
4343         by zero as VR_UNDEFINED.
4345 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
4347         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
4348         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
4349         (<sincos>mode2): New expander.
4350         (sincos_extend<mode>xf3_i387): Remove insn pattern.
4351         (sincos -> sin, cos splitters): Remove splitter patterns.
4352         (sincos<mode>3): Change operand 2 predicate to general_operand.
4353         Extend operand 2 to XFmode and generate sincosxf3 insn.
4354         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
4355         Change operand 3 predicate to const1_operand.
4356         (fptan_extend<mode>xf4_i387): Remove insn pattern.
4357         (tanxf2): Update operands in the call to fptanxf4_i387.
4358         (tan<mode>2): Change operand 1 predicate to general_operand.
4359         Extend operand 1 to XFmode and generate tanxf3 insn.
4360         (atan2xf3): Rename from *fpatanxf3_i387.
4361         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
4362         (atan2xf3): Remove expander.
4363         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
4364         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
4365         (atan<mode>2): Change operand 1 predicate to general_operand.
4366         Extend operand 1 to XFmode and generate atanxf3 insn.
4368 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
4370         * config/i386/i386.md (x87/SSE constant load splitter): Use
4371         memory_operand instead of nonimmediate_operand for input operand
4372         predicate.
4374 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
4376         * config/i386/i386.md (float partial SSE register stall splitter): Move
4377         splitter near its instruction pattern.
4378         (float_extend partial SSE register stall splitter): Ditto.
4379         (float_truncate partial SSE register stall splitter): Ditto.
4381 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
4383         PR target/86794
4384         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4385         to speculation_safe_value_not_needed.
4387         PR target/85666
4388         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
4389         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
4390         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
4391         leaf_function_p, instead use has_hard_reg_initial_val.
4393 2018-09-09  Nathan Sidwell  <nathan@acm.org>
4395         * gcc.h (pfatal_with_name): Don't declare here.
4396         * gcc.c (pfatal_with_name): Make static.
4398 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
4400         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
4401         earlyclobber.
4403 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
4405         PR rtl-optimization/85458
4406         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
4407         priority hook to reduce the priority of EXPR.
4409 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
4411         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
4412         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
4413         with X87_ENABLE_FLOAT.  Remove preparation code.
4414         (*float<SWI48:mode><MODEF:mode>2): Rename from
4415         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
4416         math using "enabled" attribute.
4417         (*floatdi<MODEF:mode>2_i387): Rename from
4418         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
4419         enable for 32bit targets only.
4420         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
4421         splitter.
4422         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
4423         as operand 1 predicate.  Rewrite as define_insn_and_split.
4424         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
4426 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
4428         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
4429         to fallthru to FLOAT case.
4431 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
4433         PR target/86731
4434         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
4435         around folding of vec_sl to handle out of range shift values.
4437 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
4439         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
4440         Update callers to gen_fix_trunc<mode>_i387_fisttp
4441         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
4442         nonimmediate_operand.
4443         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
4444         and corresponding splitters.
4445         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
4446         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
4447         (fix_truncdi_i387_with_temp): Remove insn pattern
4448         and corresponding splitters.
4449         (fix_trunc<mode>_i387): Change operand 0 predicate to
4450         nonimmediate_operand.
4451         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
4452         and corresponding splitters.
4453         (*fistdi2_1): Remove.
4454         (fistdi2): Ditto.
4455         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
4456         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
4457         (*fist<mode>2_1): Remove.
4458         (fist<mode>2): Ditto.
4459         (fist<mode>2_with_temp): Remove insn pattern and corresponding
4460         splitters.
4461         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
4462         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
4463         (fistdi2_<rounding>): Change operand 0 predicate to
4464         nonimmediate_operand.
4465         (fistdi2_<rounding>_with_temp): Remove insn pattern
4466         and corresponding splitters.
4467         (fist<mode>2_<rounding>): Change operand 0 predicate to
4468         nonimmediate_operand.
4469         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
4470         and corresponding splitters.
4472         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
4474 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4476         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
4477         the init value.
4479 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
4481         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4482         early gimple folding of vec_splat().
4483         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
4484         * gimple-fold.h: Add an extern define for tree_vec_extract().
4486 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
4488         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
4489         wrappers around TREE_TYPE comparisons.
4491 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
4493         PR target/80080
4494         * config/s390/predicates.md: Add nonsym_memory_operand.
4495         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
4496         contains a SYMBOL_REF, load it into an intermediate pseudo.
4497         (s390_emit_compare_and_swap): Legitimize operand.
4498         * config/s390/s390.md: Use the new nonsym_memory_operand
4499         with UNSPECV_CAS patterns.
4501 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
4503         PR target/80080
4504         * config/s390/s390-passes.def: New file.
4505         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
4506         declaration.
4507         (make_pass_s390_early_mach): Add declaration.
4508         * config/s390/s390.c (make_pass_s390_early_mach):
4509         (s390_option_override): Remove dynamic registration.
4510         * config/s390/t-s390: Add s390-passes.def.
4512 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
4514         * config/s390/s390.c (s390_decompose_constant_pool_ref):
4515         Remove UNSPEC_LTREL_BASE check.
4516         (annotate_constant_pool_refs): Likewise.
4517         (find_constant_pool_ref): Likewise.
4518         (find_ltrel_base): Removed.
4519         (replace_ltrel_base): Removed.
4520         (s390_mainpool_finish): Remove replace_ltrel_base call.
4521         (s390_chunkify_start): Remove pending LTREL_BASE logic.
4522         (s390_chunkify_finish): Remove replace_ltrel_base call.
4523         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
4525 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
4527         PR target/86779
4528         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
4529         to speculation_safe_value_not_needed.
4531 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
4532             Bernd Schmidt  <bernds_cb1@t-online.de>
4534         * config/nvptx/nvptx-opts.h: New file.
4535         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
4536         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
4537         (ASM_SPEC): Define.
4538         (TARGET_SM35): New macro.
4539         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
4540         correct predicate.
4541         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
4542         values.
4543         (misa=): New option.
4544         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
4546 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
4548         * config/i386/i386.md (truncdfsf2): Remove expander.
4549         (truncdfsf2_with_temp): Ditto.
4550         (truncxf<mode>2): Ditto.
4551         (*truncdfsf_fast_mixed): Remove insn pattern.
4552         (*truncdfsf_fast_i387): Ditto.
4553         (*truncdfsf_mixed): Ditto.
4554         (*truncdfsf_i387): Ditto.
4555         (*truncdfsf2_i387_1): Ditto.
4556         (*truncxfsf2_mixed): Ditto.
4557         (*truncxfdf2_mixed): Ditto.
4558         (*truncxf<mode>2_i387_noop): Ditto. Update callers
4559         to call gen_truncxf<mode>2 instead.
4560         (*truncxf<mode>2_i387): Remove.
4561         (reg->reg splitters): Remove splitter pattern.
4562         (reg->mem splitters): Ditto.
4564         (truncdfsf2): New insn pattern.
4565         (truncxf<mode>2): Ditto.
4567 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4569         * tree-ssa-math-opts.c (is_mult_by): New function.
4570         (is_square_of): Use the above.
4571         (optimize_recip_sqrt): New function.
4572         (pass_cse_reciprocals::execute): Use the above.
4574 2018-09-05  Richard Biener  <rguenther@suse.de>
4576         PR bootstrap/87134
4577         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
4578         to zero-init the emplaced vec.
4580 2018-09-05  Martin Liska  <mliska@suse.cz>
4582         PR tree-optimization/87205
4583         * tree-switch-conversion.c (pass_lower_switch::execute):
4584         Group cases for switch statements.
4586 2018-09-05  Richard Biener  <rguenther@suse.de>
4588         PR tree-optimization/87217
4589         * tree-ssa-sccvn.c (vuse_valueize): New.
4590         (vn_reference_lookup_pieces): Use it.
4591         (vn_reference_lookup): Likewise.
4593 2018-09-05  Nathan Sidwell  <nathan@acm.org>
4595         PR c++/87137
4596         * stor-layout.c (place_field): Scan forwards to check last
4597         bitfield when ms_bitfield_placement is in effect.
4599 2018-09-05  Richard Biener  <rguenther@suse.de>
4601         PR bootstrap/87225
4602         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
4603         return.
4605 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
4606             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
4608         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
4609         * config.gcc (extra_objs): Build it.
4610         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
4611         Likewise.
4612         * config/aarch64/aarch64-passes.def
4613         (pass_tag_collision_avoidance): New pass.
4614         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
4615         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
4616         (aarch64_classify_address): Remove static qualifier.
4617         (aarch64_address_info, aarch64_address_type): Move to...
4618         * config/aarch64/aarch64-protos.h: ... here.
4619         (make_pass_tag_collision_avoidance): New function.
4620         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
4621         New tuning flag.
4623 2018-09-05  Martin Liska  <mliska@suse.cz>
4625         * doc/gcov.texi: Update documentation of humar
4626         readable mode.
4627         * gcov.c (format_count): Print one decimal place, it provides
4628         more fine number of situations like '1G' vs. '1.4G'.
4630 2018-09-05  Martin Liska  <mliska@suse.cz>
4632         PR target/87164
4633         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
4634         * optc-gen.awk: Allow 'Var' for Deprecated options in order
4635         to generate a MASK value.
4637 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
4639         PR debug/86593
4640         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
4641         if frame pointer isn't used.
4642         (compute_frame_pointer_to_fb_displacement): Likewise.
4644 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
4646         PR target/87198
4647         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
4648         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
4649         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
4650         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
4651         and OPTION_MASK_ISA_XSAVEC_UNSET.
4653 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
4655         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
4656         XOR operations in NAND case.
4658 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
4660         * wide-int-range.cc (wide_int_range_convert): New.
4661         * wide-int-range.h (wide_int_range_convert): New.
4662         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
4663         code into wide_int_range_convert.
4664         (extract_range_into_wide_ints): Do not munge anti range constants
4665         into the entire domain.  Just return the range back.
4667 2018-09-04  Martin Liska  <mliska@suse.cz>
4669         * genmatch.c (output_line_directive): Add new argument
4670         fnargs.
4671         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
4673 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
4675         * doc/invoke.texi (Option Summary): Add whitespace.
4677         * doc/invoke.texi (Option Summary): Add -Waligned-new.
4679 2018-09-04  Richard Biener  <rguenther@suse.de>
4681         PR tree-optimization/87211
4682         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
4683         backedge value we're supposed to treat as VARYING also number
4684         the PHI to VARYING in case it got a different value-number already.
4686 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
4688         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
4689         (extract_range_from_binary_expr_1): Do not call
4690         vrp_can_optimize_bit_op.
4691         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
4692         static.
4693         (wide_int_range_get_mask_and_bounds): New.
4694         (wide_int_range_optimize_bit_op): New.
4695         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
4696         (wide_int_range_bit_and): Same.
4697         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
4698         (wide_int_range_optimize_bit_op): New.
4699         (wide_int_range_get_mask_and_bounds): New.
4701 2018-09-04  Richard Biener  <rguenther@suse.de>
4703         PR tree-optimization/87176
4704         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
4705         variable.  When value-numbering a virtual PHI node make sure
4706         to not value-number to the backedge value.
4708 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
4710         * doc/extend.texi (Long Long, Hex Floats): Document support for
4711         long long and hex floats in more recent versions of ISO C++.
4713 2018-09-03  Richard Biener  <rguenther@suse.de>
4715         PR tree-optimization/87177
4716         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
4717         cleanup.
4719 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4721         * bb-reorder.c (edge_order): Convert to C-qsort-style
4722         tri-state comparator.
4723         (reorder_basic_blocks_simple): Change std::stable_sort to
4724         gcc_stablesort.
4726 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4728         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
4729         tri-state comparator.
4730         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
4732 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4734         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
4735         (mergesort): ... here as maximum count for using netsort.
4736         (gcc_qsort): Set nlim to 3 if stable sort is requested.
4737         (gcc_stablesort): New.
4738         * system.h (gcc_stablesort): Declare.
4740 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
4742         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
4743         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
4744         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
4746 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
4748         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
4749         lxsdx and stxsdx alternatives.
4750         (*mov<mode>_hardfloat64): Ditto.
4751         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
4753 2018-09-03  Richard Biener  <rguenther@suse.de>
4755         PR tree-optimization/87200
4756         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
4757         simplify result.
4759 2018-09-03  Martin Liska  <mliska@suse.cz>
4761         PR tree-optimization/87201
4762         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
4763         Fix parenthesis in an expression.
4765 2018-09-03  Richard Biener  <rguenther@suse.de>
4767         PR tree-optimization/87197
4768         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
4769         visited.  CSE the VN_INFO hashtable lookup.
4771         PR tree-optimization/87169
4772         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
4773         iterating make sure there's no extra backedges from irreducible
4774         regions feeding the header.  Mark the destination block
4775         executable.
4777 2018-09-03  Martin Liska  <mliska@suse.cz>
4779         PR driver/83193
4780         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
4781         * common/common-targhooks.c (default_get_valid_option_values):
4782         New function.
4783         * common/common-targhooks.h (default_get_valid_option_values):
4784         Likewise.
4785         * common/config/i386/i386-common.c: Move processor_target_table
4786         from i386.c.
4787         (ix86_get_valid_option_values): New function.
4788         (TARGET_GET_VALID_OPTION_VALUES): New macro.
4789         * config/i386/i386.c (struct ptt): Move to i386-common.c.
4790         (PTA_*): Move all defined masks into i386-common.c.
4791         (ix86_function_specific_restore): Use new processor_cost_table.
4792         * config/i386/i386.h (struct ptt): Moved from i386.c.
4793         (struct pta): Likewise.
4794         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
4795         * doc/tm.texi.in: Likewise.
4796         * opt-suggestions.c (option_proposer::suggest_option):
4797         Pass prefix to build_option_suggestions.
4798         (option_proposer::get_completions): Likewise.
4799         (option_proposer::build_option_suggestions): Use the new target
4800         hook.
4801         * opts.c (struct option_help_tuple): New struct.
4802         (print_filtered_help): Use the new target hook.
4804 2018-09-03  Martin Liska  <mliska@suse.cz>
4806         PR middle-end/59521
4807         * predict.c (set_even_probabilities): Add likely_edges
4808         argument and handle cases where we have precisely one
4809         likely edge.
4810         (combine_predictions_for_bb): Catch also likely_edges.
4811         (tree_predict_by_opcode): Handle gswitch statements.
4812         * tree-cfg.h (find_case_label_for_value): New declaration.
4813         (find_taken_edge_switch_expr): Likewise.
4814         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
4815         Find pivot in decision tree based on probabily, not by number of
4816         nodes.
4818 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
4820         * doc/standards.texi (Standards): Update Objective-C reference.
4822 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
4824         * doc/install.texi (Prerequisites): Update link for MPC.
4826 2018-09-01  Michael Matz  <matz@suse.de>
4828         PR tree-optimization/87074
4829         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
4830         PHIs for outer-loop uses.
4832 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
4834         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
4835         * doc/invoke.texi (C Dialect Options): Ditto.
4837 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
4839         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
4841 2018-08-31  Richard Biener  <rguenther@suse.de>
4843         PR tree-optimization/87168
4844         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
4845         (rpo_elim::eliminate_avail): When OP was not visited it must
4846         be available.
4848 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
4850         * tree-vrp.c (copy_value_range): Convert param "from" from
4851         "value_range *" to "const value_range *".
4852         (range_is_null): Likewise for param "vr".
4853         (range_int_cst_p): Likewise.
4854         (range_int_cst_singleton_p): Likewise.
4855         (symbolic_range_p): Likewise.
4856         (value_ranges_intersect_p): Likewise for both params.
4857         (value_range_nonnegative_p): Likewise for param "vr".
4858         (value_range_constant_singleton): Likewise.
4859         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
4860         (extract_range_into_wide_ints): Likewise for param "vr".
4861         (extract_range_from_multiplicative_op): Likewise for params "vr0"
4862         and "vr1".
4863         (vrp_can_optimize_bit_op): Likewise.
4864         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
4865         "vr1_".
4866         (extract_range_from_unary_expr): Likewise.
4867         (debug_value_range): Likewise for param "vr".
4868         (value_range::dump): Add "const" qualifier.
4869         (vrp_prop::check_array_ref): Convert local "vr" from
4870         "value_range *" to "const value_range *".
4871         (vrp_prop::check_mem_ref): Likewise.
4872         (vrp_prop::visit_stmt): Likewise for local "old_vr".
4873         (vrp_intersect_ranges_1): Likewise for param "vr_1".
4874         (vrp_intersect_ranges): Likewise.
4875         (simplify_stmt_for_jump_threading): Likewise for local "vr".
4876         (vrp_prop::vrp_finalize): Likewise.
4877         * tree-vrp.h (value_range::dump): Add "const" qualifier.
4878         (vrp_intersect_ranges): Add "const" qualifier to params as above.
4879         (extract_range_from_unary_expr): Likewise.
4880         (value_range_constant_singleton): Likewise.
4881         (symbolic_range_p): Likewise.
4882         (copy_value_range): Likewise.
4883         (extract_range_from_binary_expr_1): Likewise.
4884         (range_int_cst_p): Likewise.
4885         (vrp_set_zero_nonzero_bits): Likewise.
4886         (range_int_cst_singleton_p): Likewise.
4888 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
4890         * config/aarch64/arm_neon.h (vabsd_s64): New.
4891         (vnegd_s64): Likewise.
4893 2018-08-31  Martin Jambor  <mjambor@suse.cz>
4895         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
4897 2018-08-31  Martin Liska  <mliska@suse.cz>
4899         * ipa-icf.c (sem_item::add_type): Use
4900         sem_item::m_type_hash_cache.
4901         * ipa-icf.h: Move the cache from sem_item_optimizer
4902         to sem_item.
4904 2018-08-31  Nathan Sidwell  <nathan@acm.org>
4906         * doc/extend.texi (Backwards Compatibility): Remove implicit
4907         extern C leeway of () being (...).
4909 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4911         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
4913 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
4915         PR target/86684
4916         PR target/87149
4917         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
4919 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
4921         PR middle-end/87138
4922         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
4923         gen_int_mode.  Formatting fixes.
4925 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
4927         * target.def (custom_function_descriptors): Improve documentation.
4928         * doc/tm.texi.in (Trampolines): Expand discussion of function
4929         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
4930         beginning of the section.
4931         * doc/tm.texi: Regenerated.
4933 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
4935         * cfg.h (class auto_edge_flag): Spell out the template-id of the
4936         base class in the initializer list.  This is a workaround for
4937         building with older GCC.
4938         (class auto_bb_flag): Likewise.
4940 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
4942         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
4943         (altivec_vcmpequ<VI_char>_p): Remove star.
4944         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
4945         vector load modes.
4946         (expand_strncmp_vec_sequence): New function.
4947         (emit_final_str_compare_vec): New function.
4948         (expand_strn_compare): Add support for vector strncmp.
4949         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
4950         length specification to bytes.
4951         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
4952         (vcmpnezb_p): New pattern.
4953         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
4954         for option -mstring-compare-inline-limit.
4956 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
4958         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
4959         (PTA_SKYLAKE): Add PTA_AES.
4960         (PTA_GOLDMONT): Likewise.
4962 2018-08-29  Jan Hubicka  <jh@suse.cz>
4964         PR lto/86517
4965         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
4966         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
4968 2018-08-29  Jan Hubicka  <jh@suse.cz>
4970         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
4971         TYPE_STUB_DECL.
4972         (hash_tree): Do not visit TYPE_STUB_DECL.
4973         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
4974         stream TYPE_STUB_DECL.
4975         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
4976         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
4977         after free_lang_data.
4978         (type_in_anonymous_namespace_p): Likewise.
4980 2018-08-29  Jan Hubicka  <jh@suse.cz>
4982         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
4983         comment that it has to be even number.
4984         (class sreal): Change m_sig type to int32_t.
4985         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
4986         int64_t for temporary calculations.
4987         (sreal_verify_basics): Drop one bit from minimum and maximum.
4989 2018-08-30  Richard Biener  <rguenther@suse.de>
4991         PR tree-optimization/87147
4992         * tree-ssa-sccvn.c (SSA_VISITED): New function.
4993         (visit_phi): When the degenerate result is from the backedge and
4994         we didn't visit its definition yet drop to VARYING.
4995         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
4997 2018-08-29  Jan Hubicka  <jh@suse.cz>
4999         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
5000         DECL_VINDEX.
5001         (hash_tree): Likewise.
5003 2018-08-29  Jan Hubicka  <jh@suse.cz>
5005         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
5006         and TYPE_NEXT_REF_TO.
5008 2018-08-29  Jan Hubicka  <jh@suse.cz>
5010         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
5011         comment that it has to be even number.
5012         (class sreal): Change m_sig type to int32_t.
5013         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
5014         int64_t for temporary calculations.
5015         (sreal_verify_basics): Drop one bit from minimum and maximum.
5017 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
5019         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
5021 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
5023         PR middle-end/86995
5024         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
5025         if to_add is negative.
5027 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5029         PR middle-end/87053
5030         * builtins.c (c_strlen): Improve range checks.
5032 2018-08-29  Martin Sebor  <msebor@redhat.com>
5033             Jeff Law  <law@redhat.com>
5035         PR tree-optimization/86714
5036         PR tree-optimization/86711
5037         * builtins.c (c_strlen): Add arguments to call to string_constant.
5038         * expr.c (string_constant): Add argument.  Detect missing nul
5039         terminator and outermost declaration it's missing in.
5040         * expr.h (string_constant): Add argument.
5041         * fold-const.c (read_from_constant_string): Add arguments to call to
5042         string_constant.
5043         (c_getstr): Likewise.
5044         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
5045         to string_constant.
5046         * tree-ssa-strlen.c (get_stridx): Likewise.
5048 2018-08-29  Jan Hubicka  <jh@suse.cz>
5050         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
5051         Do not stream DECL_VINDEX.
5052         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
5053         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
5054         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
5056 2018-08-29  Richard Biener  <rguenther@suse.de>
5058         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
5059         virtual operands that are not default defs to honor region
5060         boundaries.
5061         (rpo_vn_valueize): Remove ineffective code here.
5063 2018-08-29  Richard Biener  <rguenther@suse.de>
5065         PR tree-optimization/87132
5066         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
5067         when skipping defs reachable over backedges.
5069 2018-08-29  Richard Biener  <rguenther@suse.de>
5071         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
5072         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
5073         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
5074         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
5075         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
5076         (vn_reference_lookup_3): Remove use of const_parms.
5077         (free_rpo_vn): Do not free const_parms.
5078         (do_rpo_vn): Do not call init_const_parms.
5079         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
5080         SSA_NAME_POINTS_TO_READONLY_MEMORY.
5081         (call_may_clobber_ref_p_1): Likewise.
5083 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
5085         PR other/86726
5086         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
5087         (-O): Ditto.
5088         (-ftree-scev-cprop): Document.
5090 2018-08-29  Jan Hubicka  <jh@suse.cz>
5092         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
5093         parameters.
5094         (sreal constructor): Update.
5095         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
5096         sreal:operator/): Update.
5098 2018-08-29  Martin Liska  <mliska@suse.cz>
5100         * tree-switch-conversion.c (switch_conversion::expand):
5101         Strenghten assumption about gswitch statements.
5103 2018-08-29  Richard Biener  <rguenther@suse.de>
5105         PR tree-optimization/87117
5106         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
5107         re-value-number released SSA VDEFs.
5109 2018-08-29  Richard Biener  <rguenther@suse.de>
5111         PR tree-optimization/87126
5112         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
5114 2018-08-28  Jim Wilson  <jimw@sifive.com>
5116         * config/riscv/pic.md: Rewrite.
5117         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
5118         invalid address.
5119         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
5120         (SOFTF, default_load, softload, softstore): New.
5122 2018-08-28  Jeff Law  <law@redhat.com>
5124         * fold-const.c (fold_binary_loc): Remove recently added assert.
5126 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
5128         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
5129         to OP parmeter of generated function.
5131 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
5133         PR tree-optimization/87009
5134         * match.pd: Add boolean optimizations.
5136 2018-08-28  Martin Sebor  <msebor@redhat.com>
5138         PR middle-end/86631
5139         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
5140         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
5141         (pass_walloca::gate): Use it.
5142         (alloca_call_type): Same.
5143         (pass_walloca::execute): Same.
5144         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
5146 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
5148         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
5149         GCC_VERSION for usage of "__gcc_dump_printf__" format from
5150         >= 3005 to >= 9000.
5152 2018-08-28  Richard Biener  <rguenther@suse.de>
5154         PR tree-optimization/87124
5155         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
5156         constants before looking up avail.
5158 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
5160         PR middle-end/87099
5161         * calls.c (maybe_warn_nonstring_arg): Punt early if
5162         warn_stringop_overflow is zero.  Don't call get_range_strlen
5163         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
5164         Swap comparison operands to have constants on rhs.  Only use
5165         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
5166         increment lenrng[0].
5168 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
5170         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
5171         use of tree_to_shwi.  Remove duplicated test for the size being
5172         a whole number of bytes.
5174 2018-08-28  Richard Biener  <rguenther@suse.de>
5176         PR tree-optimization/87117
5177         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
5178         Handle removed stmt without LHS (GIMPLE_NOP).
5180 2018-08-28  Richard Biener  <rguenther@suse.de>
5182         PR tree-optimization/87117
5183         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
5184         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
5186 2018-08-28  Richard Biener  <rguenther@suse.de>
5188         PR tree-optimization/87117
5189         * tree-ssa-pre.c (compute_avail): Do not make expressions
5190         with predicated values available.
5191         (get_expr_value_id): Assert we do not run into predicated value
5192         expressions.
5194 2018-08-28  Richard Biener  <rguenther@suse.de>
5196         PR tree-optimization/87117
5197         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
5198         get virtual operands.
5199         (get_expr_operands): Handle STRING_CST like other decls.
5201 2018-08-28  Martin Liska  <mliska@suse.cz>
5203         * tree.h: Update documentation of fndecl_built_in_p
5204         functions.
5207 2018-08-27  Jeff Law  <law@redhat.com>
5208         PR tree-optimization/87110
5209         * tree-ssa-dse.c (compute_trims): Handle non-constant
5210         TYPE_SIZE_UNIT.
5212 2018-08-27  Martin Sebor  <msebor@redhat.com>
5214         PR tree-optimization/86914
5215         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
5217 2018-08-27  Martin Sebor  <msebor@redhat.com>
5219         PR tree-optimization/87112
5220         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
5221         the type of the bound argument.
5223 2018-08-27  Jeff Law  <law@redhat.com>
5225         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
5226         type does not have a TYPE_SIZE_UNIT.
5228 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
5230         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
5231         with include of backend.h.
5233 2018-08-27  Richard Biener  <rguenther@suse.de>
5235         PR tree-optimization/86927
5236         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
5237         use const cond reduction code.
5239 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
5241         PR tree-optimization/85758
5242         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
5244 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
5246         PR c++/87091
5247         * diagnostic-show-locus.c (class layout_range): Update for
5248         conversion of show_caret_p to a tri-state.
5249         (layout_range::layout_range): Likewise.
5250         (make_range): Likewise.
5251         (layout::maybe_add_location_range): Likewise.
5252         (layout::should_print_annotation_line_p): Don't show annotation
5253         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
5254         (layout::get_state_at_point): Update for conversion of
5255         show_caret_p to a tri-state.  Bail out early for
5256         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
5257         underlining or source colorization.
5258         (gcc_rich_location::add_location_if_nearby): Update for conversion
5259         of show_caret_p to a tri-state.
5260         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
5261         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
5262         Likewise.
5263         (selftest::test_one_liner_labels): Likewise.
5264         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
5265         conversion of show_caret_p to a tri-state.
5266         * pretty-print.c (text_info::set_location): Likewise.
5267         * pretty-print.h (text_info::set_location): Likewise.
5268         * substring-locations.c (format_warning_n_va): Likewise.
5269         * tree-diagnostic.c (default_tree_printer): Likewise.
5270         * tree-pretty-print.c (newline_and_indent): Likewise.
5272 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
5274         PR c++/87091
5275         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
5276         line above for line-insertion fix-it hints.
5277         (selftest::test_fixit_insert_containing_newline): Update the
5278         expected results, and add a test with line-numbering enabled.
5280 2018-08-27  Martin Liska  <mliska@suse.cz>
5282         PR sanitizer/86962
5283         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
5284         params with DECL_HAS_VALUE_EXPR_P.
5286 2018-08-27  Martin Liska  <mliska@suse.cz>
5288         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
5289         selected expansion strategy.
5291 2018-08-27  Martin Liska  <mliska@suse.cz>
5293         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
5294         * builtins.c (is_builtin_fn): Likewise.
5295         * attribs.c (diag_attr_exclusions): Use new function
5296         fndecl_built_in_p and remove check for FUNCTION_DECL if
5297         possible.
5298         (builtin_mathfn_code): Likewise.
5299         (fold_builtin_expect): Likewise.
5300         (fold_call_expr): Likewise.
5301         (fold_builtin_call_array): Likewise.
5302         (fold_call_stmt): Likewise.
5303         (set_builtin_user_assembler_name): Likewise.
5304         (is_simple_builtin): Likewise.
5305         * calls.c (gimple_alloca_call_p): Likewise.
5306         (maybe_warn_nonstring_arg): Likewise.
5307         * cfgexpand.c (expand_call_stmt): Likewise.
5308         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
5309         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
5310         (cgraph_node::verify_node): Likewise.
5311         * cgraphclones.c (build_function_decl_skip_args): Likewise.
5312         (cgraph_node::create_clone): Likewise.
5313         * config/arm/arm.c (arm_insert_attributes): Likewise.
5314         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
5315         * dse.c (scan_insn): Likewise.
5316         * expr.c (expand_expr_real_1): Likewise.
5317         * fold-const.c (operand_equal_p): Likewise.
5318         (fold_binary_loc): Likewise.
5319         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
5320         * gimple-low.c (lower_stmt): Likewise.
5321         * gimple-pretty-print.c (dump_gimple_call): Likewise.
5322         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
5323         Likewise.
5324         * gimple.c (gimple_build_call_from_tree): Likewise.
5325         (gimple_call_builtin_p): Likewise.
5326         (gimple_call_combined_fn): Likewise.
5327         * gimplify.c (gimplify_call_expr): Likewise.
5328         (gimple_boolify): Likewise.
5329         (gimplify_modify_expr): Likewise.
5330         (gimplify_addr_expr): Likewise.
5331         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
5332         * ipa-cp.c (determine_versionability): Likewise.
5333         * ipa-fnsummary.c (compute_fn_summary): Likewise.
5334         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
5335         * ipa-split.c (visit_bb): Likewise.
5336         (split_function): Likewise.
5337         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
5338         * lto-cgraph.c (input_node): Likewise.
5339         * lto-streamer-out.c (write_symbol): Likewise.
5340         * omp-low.c (setjmp_or_longjmp_p): Likewise.
5341         (lower_omp_1): Likewise.
5342         * predict.c (strip_predict_hints): Likewise.
5343         * print-tree.c (print_node): Likewise.
5344         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
5345         * trans-mem.c (is_tm_irrevocable): Likewise.
5346         (is_tm_load): Likewise.
5347         (is_tm_simple_load): Likewise.
5348         (is_tm_store): Likewise.
5349         (is_tm_simple_store): Likewise.
5350         (is_tm_abort): Likewise.
5351         (tm_region_init_1): Likewise.
5352         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
5353         * tree-cfg.c (verify_gimple_call): Likewise.
5354         (move_stmt_r): Likewise.
5355         (stmt_can_terminate_bb_p): Likewise.
5356         * tree-eh.c (lower_eh_constructs_2): Likewise.
5357         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
5358         * tree-inline.c (remap_gimple_stmt): Likewise.
5359         (copy_bb): Likewise.
5360         (estimate_num_insns): Likewise.
5361         (fold_marked_statements): Likewise.
5362         * tree-sra.c (scan_function): Likewise.
5363         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
5364         (optimize_stack_restore): Likewise.
5365         (pass_fold_builtins::execute): Likewise.
5366         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
5367         (mark_all_reaching_defs_necessary_1): Likewise.
5368         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
5369         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
5370         (pass_forwprop::execute): Likewise.
5371         * tree-ssa-loop-im.c (stmt_cost): Likewise.
5372         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
5373         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
5374         * tree-ssa-strlen.c (get_string_length): Likewise.
5375         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
5376         (find_func_aliases_for_call): Likewise.
5377         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
5378         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
5379         * tree-tailcall.c (find_tail_calls): Likewise.
5380         * tree.c (need_assembler_name_p): Likewise.
5381         (free_lang_data_in_decl): Likewise.
5382         (get_call_combined_fn): Likewise.
5383         * ubsan.c (is_ubsan_builtin_p): Likewise.
5384         * varasm.c (incorporeal_function_p): Likewise.
5385         * tree.h (DECL_BUILT_IN): Remove and replace with
5386         fndecl_built_in_p.
5387         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
5388         (fndecl_built_in_p): New.
5390 2018-08-27  Martin Liska  <mliska@suse.cz>
5392         PR tree-optimization/86847
5393         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
5394         Dump also subtree probability.
5395         (switch_decision_tree::do_jump_if_equal): New function.
5396         (switch_decision_tree::emit_case_nodes): Handle special
5397         situations in balanced tree that can be emitted much simpler.
5398         Fix calculation of probabilities that happen in tree expansion.
5399         * tree-switch-conversion.h (struct cluster): Add
5400         is_single_value_p.
5401         (struct simple_cluster): Likewise.
5402         (struct case_tree_node): Add new function has_child.
5403         (do_jump_if_equal): New.
5405 2018-08-27  Martin Liska  <mliska@suse.cz>
5407         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
5408         Add new argument to bit_test_cluster constructor.
5409         (bit_test_cluster::emit): Set bits really number of values
5410         handlel by a test.
5411         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
5412         probability argument.
5413         * tree-switch-conversion.h (struct bit_test_cluster):
5414         Add m_handles_entire_switch.
5416 2018-08-27  Martin Liska  <mliska@suse.cz>
5418         PR tree-optimization/86702
5419         * tree-switch-conversion.c (jump_table_cluster::emit):
5420         Make probabilities even for values in jump table
5421         according to number of cases handled.
5422         (switch_decision_tree::compute_cases_per_edge): Pass
5423         argument to reset_out_edges_aux function.
5424         (switch_decision_tree::analyze_switch_statement): Likewise.
5425         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
5426         Make it static.
5428 2018-08-27  Martin Liska  <mliska@suse.cz>
5430         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
5431         cfun argument explicitly.
5432         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
5433         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
5434         function gimple_switch_default_bb.
5435         (convert_switch_statements):
5436         (expand_builtins):
5437         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
5438         * stmt.c (label_to_block_fn): Use label_to_block and pass
5439         cfun argument explicitly and use gimple_switch_label_bb.
5440         (expand_case): Likewise.
5441         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
5442         cfun argument explicitly. Likewise.
5443         (make_edges_bb): Likewise.
5444         (make_cond_expr_edges): Likewise.
5445         (get_cases_for_edge): Likewise.
5446         (make_gimple_switch_edges): Likewise.
5447         (label_to_block_fn): Likewise.
5448         (label_to_block): Likewise.
5449         (make_goto_expr_edges): Likewise.
5450         (make_gimple_asm_edges): Likewise.
5451         (main_block_label): Likewise.
5452         (group_case_labels_stmt): Likewise.
5453         (find_taken_edge_computed_goto): Likewise.
5454         (find_taken_edge_switch_expr): Likewise.
5455         (gimple_verify_flow_info): Likewise.
5456         (gimple_redirect_edge_and_branch): Likewise.
5457         (gimple_switch_label_bb): New function.
5458         (gimple_switch_default_bb): Likewise.
5459         (gimple_switch_edge): Likewise.
5460         (gimple_switch_default_edge): Likewise.
5461         * tree-cfg.h (label_to_block_fn): Remove and replace ...
5462         (label_to_block): ... with this.
5463         (gimple_switch_label_bb): New.
5464         (gimple_switch_default_bb): Likewise.
5465         (gimple_switch_edge): Likewise.
5466         (gimple_switch_default_edge): Likewise.
5467         * tree-cfgcleanup.c (convert_single_case_switch): Use
5468         new gimple functions and pass new argument to label_to_block.
5469         (cleanup_control_flow_bb):
5470         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
5471         cfun argument explicitly.
5472         (make_eh_edges): Likewise.
5473         (redirect_eh_dispatch_edge): Likewise.
5474         (lower_resx): Likewise.
5475         (lower_eh_dispatch): Likewise.
5476         (maybe_remove_unreachable_handlers): Likewise.
5477         (unsplit_eh): Likewise.
5478         (cleanup_empty_eh): Likewise.
5479         (verify_eh_edges): Likewise.
5480         (verify_eh_dispatch_edge): Likewise.
5481         * tree-ssa-dom.c (record_edge_info): Likewise.
5482         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
5483         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
5484         (thread_through_normal_block): Likewise.
5485         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
5486         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
5487         * tree-switch-conversion.c (switch_conversion::collect): Use new
5488         gimple functions.
5489         (switch_conversion::check_final_bb): Likewise.
5490         (switch_conversion::gather_default_values): Pass new argument
5491         to label_to_block.
5492         (switch_conversion::build_constructors): Likewise.
5493         (switch_decision_tree::compute_cases_per_edge): Use new
5494         gimple_switch_edge function.
5495         (switch_decision_tree::analyze_switch_statement): Pass new argument
5496         to label_to_block.
5497         (switch_decision_tree::try_switch_expansion): Use
5498         gimple_switch_default_edge.
5499         * tree-vrp.c (find_switch_asserts): Pass new argument
5500         to label_to_block.
5501         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
5502         (vr_values::simplify_switch_using_ranges): Likewise.
5504 2018-08-27  Richard Biener  <rguenther@suse.de>
5506         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
5507         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
5509         * tree-ssa-sccvn.h (struct vn_pval): New structure.
5510         (struct vn_nary_op_s): Add unwind_to member.  Add
5511         predicated_values flag and put result into a union together
5512         with a linked list of vn_pval.
5513         (struct vn_ssa_aux): Add name member to make maintaining
5514         a map of SSA name to vn_ssa_aux possible.  Remove no longer
5515         needed info, dfsnum, low, visited, on_sccstack, use_processed
5516         and range_info_anti_range_p members.
5517         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
5518         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
5519         New functions.
5520         (vn_valueize): New global.
5521         (vn_context_bb): Likewise.
5522         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
5523         VN_INFO_PTR_INFO): Remove.
5524         * tree-ssa-sccvn.c: ... (rewrite)
5525         (pass_fre::execute): For -O2+ initialize loops and run
5526         RPO VN in optimistic mode (iterating).  For -O1 and -Og
5527         run RPO VN in non-optimistic mode.
5528         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
5529         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
5530         * doc/invoke.texi (sccvn-max-scc-size): Remove.
5531         (rpo-vn-max-loop-depth): Document.
5532         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
5533         when valuezing the VUSE signals we walked out of the region.
5534         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
5535         (phi_translate): Set VN context block to use for availability
5536         lookup.
5537         (compute_avail): Likewise.
5538         (pre_valueize): New function.
5539         (pass_pre::execute): Adjust to the RPO VN API.
5541         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
5542         (propagate_constants_for_unrolling): Remove.
5543         (tree_unroll_loops_completely): Perform value-numbering
5544         on the unrolled bodies loop parent.
5546 2018-08-27  Richard Biener  <rguenther@suse.de>
5548         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
5549         for partial antic compute.
5551 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
5553         PR rtl-optimization/87065
5554         * combine.c (simplify_if_then_else): Formatting fix.
5555         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
5556         check.
5557         (known_cond): Don't return const_true_rtx for vector modes.  Use
5558         CONST0_RTX instead of const0_rtx.  Formatting fixes.
5560 2018-08-27  Martin Liska  <mliska@suse.cz>
5562         PR gcov-profile/87069
5563         * gcov.c (process_file): Record files already processed
5564         and warn about a file being processed multiple times.
5566 2018-08-27  Martin Liska  <mliska@suse.cz>
5568         PR driver/83193
5569         * config/aarch64/aarch64.c (aarch64_override_options_internal):
5570         Set default values for x_aarch64_*_string strings.
5571         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
5572         prefix.  For -mabi do not print '=ABI' in help and use
5573         <option_value> format for -msve-vector-bits and -moverride
5574         options.
5576 2018-08-26  Jeff Law  <law@redhat.com>
5578         * config/mips/frame-header-opt.c: Include "backend.h" rather than
5579         "cfg.h"
5581 2018-08-26  Marek Polacek  <polacek@redhat.com>
5583         PR c++/87029, Implement -Wredundant-move.
5584         * doc/invoke.texi: Document -Wredundant-move.
5586 2018-08-25  Martin Sebor  <msebor@redhat.com>
5588         PR tree-optimization/87059
5589         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
5590         to the same type as the other.
5591         * fold-const.c (fold_binary_loc): Assert expectation.
5593 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
5595         * config/darwin.c (machopic_legitimize_pic_address): Clean up
5596         extraneous parentheses, dead code section and formatting.
5598 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
5600         PR c++/87091
5601         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
5602         wide enough for jumps in the line-numbering to be visible.
5603         (layout::print_gap_in_line_numbering): New member function.
5604         (layout::calculate_line_spans): When using line numbering, merge
5605         line spans that are only 1 line apart.
5606         (diagnostic_show_locus): When printing line numbers, show gaps in
5607         line numbering directly, rather than printing headers.
5608         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
5609         line-numbering with multiple line spans.
5610         (selftest::test_fixit_insert_containing_newline_2): Add test of
5611         line-numbering, in which the spans are close enough to be merged.
5613 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
5615         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
5616         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
5617         * tree-vrp.c (range_is_nonnull): Remove.
5618         (range_includes_zero_p): Accept value_range instead of min/max.
5619         (extract_range_from_binary_expr_1): Do not early bail on
5620         POINTER_PLUS_EXPR.
5621         Use range_includes_zero_p instead of range_is_nonnull.
5622         (extract_range_from_unary_expr): Use range_includes_zero_p instead
5623         of range_is_nonnull.
5624         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
5625         special case VR_ANTI_RANGE.
5626         (vrp_finalize): Same.
5627         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
5628         instead of min/max.
5629         (range_is_nonnull): Remove.
5630         * vr-values.c (vrp_stmt_computes_nonzero): Use
5631         range_includes_zero_p instead of range_is_nonnull.
5632         (extract_range_basic): Pass value_range to range_includes_zero_p
5633         instead of range_is_nonnull.
5635 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
5637         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
5638         * emit-rtl.h (rtl_data): Remove return_bnd.
5639         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
5640         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
5641         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
5642         (POINTER_BOUNDS_MODE): Remove definition.
5643         (make_pointer_bounds_mode): Remove.
5644         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
5645         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
5646         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
5647         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
5648         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
5649         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
5650         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
5652         * config/i386/i386-modes.def (BND32, BND64): Remove.
5653         * config/i386/i386.c (dbx_register_map): Remove bound registers.
5654         (dbx64_register_map): Ditto.
5655         (svr4_dbx_register_map): Ditto.
5656         (indirect_thunk_bnd_needed): Remove.
5657         (indirect_thunks_bnd_used): Ditto.
5658         (indirect_return_bnd_needed): Ditto.
5659         (indirect_return_via_cx_bnd): Ditto.
5660         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
5661         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
5662         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
5663         (output_indirect_thunk_function): Remove handling of
5664         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
5665         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
5666         (ix86_save_reg): Remove handling of crtl->return_bnd.
5667         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
5668         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
5669         and UNSPEC_BNDLX_ADDR.
5670         (ix86_output_indirect_branch_via_reg): Remove handling of
5671         indirect_thunk_prefix_bnd.
5672         (ix86_output_indirect_branch_via_push): Ditto.
5673         (ix86_output_function_return): Ditto.
5674         (ix86_output_indirect_function_return): Ditto.
5675         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
5676         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
5677         (CALL_USED_REGISTERS): Ditto.
5678         (REG_ALLOC_ORDER): Update for removal of bound registers.
5679         (HI_REGISTER_NAMES): Ditto.
5680         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
5681         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
5682         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
5683         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
5684         (FIRST_PSEUDO_REG): Update.
5685         (BND): Remove mode iterator.
5686         * config/i386/predicates.md (bnd_mem_operator): Remove.
5688 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
5690         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
5691         vectors.
5693 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
5695         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
5696         the case in which the permute needs only a single element and
5697         repeats for every vector of the result.  Extend that case to
5698         handle variable-length vectors.
5699         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
5701 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
5703         PR debug/79342
5704         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
5705         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
5707 2018-08-24  Richard Biener  <rguenther@suse.de>
5709         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
5710         bb_flags_allocated members.
5711         (auto_flag): New RAII class for allocating flags.
5712         (auto_edge_flag): New RAII class for allocating edge flags.
5713         (auto_bb_flag): New RAII class for allocating bb flags.
5714         * cfgloop.c (verify_loop_structure): Allocate temporary edge
5715         flag dynamically.
5716         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
5717         in favor of temporarily allocated BB flag.
5718         * hsa-brig.c: Re-order includes.
5719         * hsa-dump.c: Likewise.
5720         * hsa-regalloc.c: Likewise.
5721         * print-rtl.c: Likewise.
5722         * profile-count.c: Likewise.
5724 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
5726         PR target/86989
5727         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
5728         the TOC register.
5730 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
5732         PR 87073/bootstrap
5733         * wide-int-range.cc (wide_int_range_div): Do not ignore result
5734         from wide_int_range_multiplicative_op.
5736 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5738         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
5739         "permutaion".
5741 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
5743         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
5744         to 'expanded'.
5746 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
5748         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
5749         full GENERIC expression used for replacement.
5751 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
5753         * tree-vrp.c (abs_extent_range): Remove.
5754         (extract_range_into_wide_ints): Pass wide ints by reference.
5755         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
5756         Pass wide ints by reference in all calls to
5757         extract_range_into_wide_ints.
5758         * wide-int-range.cc (wide_int_range_div): New.
5759         * wide-int-range.h (wide_int_range_div): New.
5760         (wide_int_range_includes_zero_p): New.
5761         (wide_int_range_zero_p): New.
5763 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
5765         * config/aarch64/aarch64.md (arches): New enum.
5766         (arch): New enum attr.
5767         (arch_enabled): New attr.
5768         (enabled): Now uses arch_enabled only.
5769         (simd, sve, fp16): Removed attribute.
5770         (fp): Attr now defined in terms of 'arch'.
5771         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
5772         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
5773         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
5774         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
5775         attributes into 'arch'.
5776         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
5777         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
5778         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
5779         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
5780         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
5781         'simd' attribute into 'arch'.
5782         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
5783         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
5784         Convert use of 'fp' attribute to 'arch'.
5785         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
5786         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
5787         into 'arch'.
5788         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
5789         (different modes) Merge 'fp' and 'simd' into 'arch'.
5790         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
5791         'simd' into 'arch'.
5793 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
5795         PR rtl-optimization/87026
5796         * expmed.c (canonicalize_comparison): If we can no longer create
5797         pseudoregisters, don't.
5799 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
5801         PR target/86951
5802         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
5803         prototype.
5804         * config/arm/arm.c (speculation_barrier_libfunc): New static
5805         variable.
5806         (arm_init_libfuncs): Initialize it.
5807         (arm_emit_speculation_barrier): New function.
5808         * config/arm/arm.md (speculation_barrier): Call
5809         arm_emit_speculation_barrier for architectures that do not have
5810         DSB or ISB.
5811         (speculation_barrier_insn): Only match on Armv7 or later.
5813 2018-08-23  Richard Biener  <rguenther@suse.de>
5815         PR middle-end/87024
5816         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
5817         calls.
5819 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
5821         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
5822         of single-vector TBLs.
5823         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
5824         one input is given.
5826 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
5828         PR target/85910
5829         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
5830         aarch64_evpc_tbl guard.
5832 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5834         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
5835         behaviour.
5837 2018-08-22  Martin Sebor  <msebor@redhat.com>
5839         PR middle-end/87052
5840         * tree-pretty-print.c (pretty_print_string): Add argument.
5841         (dump_generic_node): Call to pretty_print_string with string size.
5843 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
5845         PR rtl-optimization/86771
5846         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
5847         of two SETs into those two SETs, one to be placed at i2, if that SETs
5848         destination is modified between i2 and i3.
5850 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5852         PR tree-optimization/86725
5853         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
5854         function.
5855         (vect_analyze_scalar_cycles_1): Check it.
5857 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5859         PR tree-optimization/86725
5860         * tree-vect-loop.c (vect_is_simple_reduction): When treating
5861         an outer loop phi as a double reduction, make sure that the
5862         single user of the phi result is an inner loop phi.
5864 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5866         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
5867         grouped stores with gaps to a strided group.
5869 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
5871         * tree-vect-stmts.c (get_group_load_store_type)
5872         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
5873         first statement in a group.
5875 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5877         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
5878         the sequence used in gcc/gcc.c.
5880 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5882         PR other/704
5883         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
5884         building it.
5886 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5888         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
5889         Darwin10-specific unwinder-shim.
5890         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
5891         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
5892         New to cater for Darwin10 Rosetta.
5894 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5896         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
5897         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
5899 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
5901         PR bootstrap/81033
5902         PR target/81733
5903         PR target/52795
5904         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
5905         (dwarf2out_switch_text_section): Generate a local label for the
5906         second function sub-section and apply it as the second FDE start
5907         label.
5908         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
5909         second sub-section start.
5911 2018-08-22  Richard Biener  <rguenther@suse.de>
5913         PR tree-optimization/86988
5914         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
5916 2018-08-22  Richard Biener  <rguenther@suse.de>
5918         PR tree-optimization/86945
5919         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
5921 2018-08-22  Alexandre Oliva <oliva@adacore.com>
5923         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
5924         a comment about how uses of r2 for .sdata2 come about.
5926 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
5928         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
5930 2018-08-21  Marek Polacek  <polacek@redhat.com>
5932         PR c++/86981, Implement -Wpessimizing-move.
5933         * doc/invoke.texi: Document -Wpessimizing-move.
5935 2018-08-21  Jan Hubicka  <jh@suse.cz>
5937         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
5938         * tree.h (is_redundant_typedef): Remove.
5939         * dwarf2out.c (is_redundant_typedef): Turn into static function.
5941 2018-08-21  Jan Hubicka  <jh@suse.cz>
5943         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
5944         when possible.
5946 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
5948         * expmed.c (extract_low_bits): Reject invalid subregs early.
5950 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5952         PR middle-end/86121
5953         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
5954         behaviour.
5956 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
5958         * config/vxworks.h: Guard vxworks_asm_out_constructor and
5959         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
5960         * config/vxworks.c: Likewise.
5962 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
5964         * config/vxworks.c: Set targetm.have_ctors_dtors
5965         if HAVE_INITFINI_ARRAY_SUPPORT.
5966         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
5967         if HAVE_INITFINI_ARRAY_SUPPORT.
5969 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
5971         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
5972         default search path for VxWorks < 7.
5974 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5976         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
5977         (get_format_string): Refer to c_getstr.
5979 2018-08-21  Tom de Vries  <tdevries@suse.de>
5981         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
5982         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
5983         (debuginfo_early_stop): Declare.
5984         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
5985         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
5986         (debuginfo_early_stop): New function.
5987         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
5988         and debuginfo_early_stop.
5989         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
5990         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
5991         (general_init): Call debuginfo_early_init.
5992         (finalize): Call debuginfo_fini.
5993         (do_compile): Call debuginfo_init.
5994         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
5995         -fdump-early-debug.
5996         (@item -fdump-debug, @item -fdump-earlydebug): Add.
5998 2018-08-21  Tom de Vries  <tdevries@suse.de>
6000         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
6001         flag_dump_noaddr and flag_dump_unnumbered.
6003 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
6005         * wide-int-range.cc (wide_int_range_abs): New.
6006         (wide_int_range_order_set): Rename from wide_int_range_min_max.
6007         * wide-int-range.h (wide_int_range_abs): New.
6008         (wide_int_range_min_max): New.
6009         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
6010         case to call wide_int_range_abs.
6011         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
6012         (extract_range_from_abs_expr): Delete.
6014 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
6016         PR target/87033
6017         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
6018         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
6019         for indexed loads.
6021 2018-08-20  Nathan Sidwell  <nathan@acm.org>
6022             Jeff Law <law@redhat.com>
6024         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
6025         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
6027 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
6029         PR other/84889
6030         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
6031         (decl_attributes): Likewise.
6032         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
6033         instance.
6034         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6035         * diagnostic-core.h (class auto_diagnostic_group): New class.
6036         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
6037         (diagnostic_report_diagnostic): Handle the first diagnostics within
6038         a group.
6039         (emit_diagnostic): Add auto_diagnostic_group instance.
6040         (inform): Likewise.
6041         (inform_n): Likewise.
6042         (warning): Likewise.
6043         (warning_at): Likewise.
6044         (warning_n): Likewise.
6045         (pedwarn): Likewise.
6046         (permerror): Likewise.
6047         (error): Likewise.
6048         (error_n): Likewise.
6049         (error_at): Likewise.
6050         (sorry): Likewise.
6051         (fatal_error): Likewise.
6052         (internal_error): Likewise.
6053         (internal_error_no_backtrace): Likewise.
6054         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
6055         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
6056         * diagnostic.h (struct diagnostic_context): Add fields
6057         "diagnostic_group_nesting_depth",
6058         "diagnostic_group_emission_count", "begin_group_cb",
6059         "end_group_cb".
6060         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
6061         Add auto_diagnostic_group instance(s).
6062         (find_explicit_erroneous_behavior): Likewise.
6063         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
6064         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
6065         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
6066         (gimplify_va_arg_expr): Likewise.
6067         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
6068         (HSA_SORRY_AT): Likewise.
6069         * ipa-devirt.c (compare_virtual_tables): Likewise.
6070         (warn_odr): Likewise.
6071         * multiple_target.c (expand_target_clones): Likewise.
6072         * opts-common.c (cmdline_handle_error): Likewise.
6073         * reginfo.c (globalize_reg): Likewise.
6074         * substring-locations.c (format_warning_n_va): Likewise.
6075         * tree-inline.c (expand_call_inline): Likewise.
6076         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
6077         * tree-ssa-loop-niter.c
6078         (do_warn_aggressive_loop_optimizations): Likewise.
6079         * tree-ssa-uninit.c (warn_uninit): Likewise.
6080         * tree.c (warn_deprecated_use): Likewise.
6082 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
6084         PR target/87014
6085         * config/i386/i386.md (eh_return): Always update EH return
6086         address in word_mode.
6088 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
6090         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
6091         TARGET_SPLIT_COMPLEX_ARG is defined.
6093 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6095         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
6097 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6099         PR target/86984
6100         * expr.c (expand_assignment): Assert that bitpos is positive.
6101         (store_field): Likewise
6102         (expand_expr_real_1): Make sure that bitpos is positive.
6103         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
6104         integer overflow.
6106 2018-08-20  Nathan Sidwell  <nathan@acm.org>
6108         * Makefile.in (CPP_ID_DATA_H): Delete.
6109         (CPP_INTERNAL_H): Don't add it.
6110         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
6111         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
6113 2018-08-20  Richard Biener  <rguenther@suse.de>
6115         PR tree-optimization/78655
6116         * tree-vrp.c (extract_range_from_binary_expr_1): Make
6117         pointer + offset nonnull if either operand is nonnull work.
6119 2018-08-20  Tom de Vries  <tdevries@suse.de>
6121         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
6122         unless the referenced die describes the added property using
6123         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
6124         Otherwise, add a DW_AT_location to the referenced die.
6126 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
6128         PR target/86994
6129         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
6130         register_operand when calling ix86_set_reg_reg_cost.
6131         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
6132         Set *total to 0 for operands that satisfy x86_64_immediate_operand
6133         predicate and to 1 otherwise.
6135 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
6137         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
6138         emit a diagnostic that it is not supported and reset the option.
6139         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
6140         supported and consume the option.  (ASM_FINAL_SPEC): New.
6142 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
6144         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
6145         a sentence.
6147 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
6149         C-SKY port: Documentation
6151         * doc/extend.texi (C-SKY Function Attributes): New section.
6152         * doc/invoke.texi (Option Summary): Add C-SKY options.
6153         (C-SKY Options): New section.
6154         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
6156 2018-08-17  Jojo  <jijie_rong@c-sky.com>
6157             Huibin Wang  <huibin_wang@c-sky.com>
6158             Sandra Loosemore  <sandra@codesourcery.com>
6159             Chung-Lin Tang  <cltang@codesourcery.com>
6161         C-SKY port: Backend implementation
6163         * config/csky/*: New.
6164         * common/config/csky/*: New.
6166 2018-08-17  Jojo  <jijie_rong@c-sky.com>
6167             Huibin Wang  <huibin_wang@c-sky.com>
6168             Sandra Loosemore  <sandra@codesourcery.com>
6169             Chung-Lin Tang  <cltang@codesourcery.com>
6170             Andrew Jenner  <andrew@codesourcery.com>
6172         C-SKY port: Configury
6174         * config.gcc (csky-*-*): New.
6175         * configure.ac: Add csky to targets for dwarf2 debug_line support.
6176         * configure: Regenerated.
6178 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
6180         * dump-context.h: Include "dumpfile.h".
6181         (dump_context::dump_printf_va): Convert final param from va_list
6182         to va_list *.  Convert from ATTRIBUTE_PRINTF to
6183         ATTRIBUTE_GCC_DUMP_PRINTF.
6184         (dump_context::dump_printf_loc_va): Likewise.
6185         * dumpfile.c: Include "stringpool.h".
6186         (make_item_for_dump_printf_va): Delete.
6187         (make_item_for_dump_printf): Delete.
6188         (class dump_pretty_printer): New class.
6189         (dump_pretty_printer::dump_pretty_printer): New ctor.
6190         (dump_pretty_printer::emit_items): New member function.
6191         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
6192         function.
6193         (dump_pretty_printer::emit_item): New member function.
6194         (dump_pretty_printer::stash_item): New member function.
6195         (dump_pretty_printer::format_decoder_cb): New member function.
6196         (dump_pretty_printer::decode_format): New member function.
6197         (dump_context::dump_printf_va): Reimplement in terms of
6198         dump_pretty_printer.
6199         (dump_context::dump_printf_loc_va): Convert final param from va_list
6200         to va_list *.
6201         (dump_context::begin_scope): Reimplement call to
6202         make_item_for_dump_printf.
6203         (dump_printf): Update for change to dump_printf_va.
6204         (dump_printf_loc): Likewise.
6205         (selftest::test_capture_of_dump_calls): Convert "stmt" from
6206         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
6207         with %T, %E, and %G.
6208         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
6209         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
6210         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
6211         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
6212         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
6213         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
6214         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
6215         within a dump_printf_loc call to "%wu".
6216         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
6217         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
6218         missing space after "=".
6219         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
6220         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
6221         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
6222         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
6223         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
6224         duplicate "vectorized" from message.
6226 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
6228         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
6229         polyNxK_t element's TYPE_STRING_FLAG.
6231 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
6233         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
6234         (they were unnamed before).  Fix comments.
6236 2018-08-17  Nathan Sidwell  <nathan@acm.org>
6238         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
6240 2018-08-17  Richard Biener  <rguenther@suse.de>
6242         PR tree-optimization/86841
6243         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
6245 2018-08-17  Martin Liska  <mliska@suse.cz>
6247         * common.opt: Remove Warn, Init and Report for options with
6248         Ignore/Deprecated flag. Warning is done automatically for
6249         Deprecated flags.
6250         * config/i386/i386.opt: Likewise.
6251         * config/ia64/ia64.opt: Likewise.
6252         * config/rs6000/rs6000.opt: Likewise.
6253         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
6254         Remove usage of flag_check_pointer_bounds.
6255         * lto-wrapper.c (merge_and_complain): Do not handle
6256         OPT_fcheck_pointer_bounds.
6257         (append_compiler_options): Likewise.
6258         * opt-functions.awk: Do not handle Deprecated.
6259         * optc-gen.awk: Check that Var, Report and Init are not
6260         used for an option with Ignore/Deprecated flag.
6261         * opts-common.c (decode_cmdline_option): Do not report
6262         CL_ERR_DEPRECATED.
6263         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
6264         options.
6265         * opts.h (struct cl_option): Remove cl_deprecated flag.
6266         (CL_ERR_DEPRECATED): Remove error enum value.
6268 2018-08-17  Richard Biener  <rguenther@suse.de>
6270         PR middle-end/86505
6271         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
6272         across a va-arg-pack using call adjust its return value accordingly.
6274 2018-08-16  Martin Sebor  <msebor@redhat.com>
6276         PR tree-optimization/86853
6277         * gimple-ssa-sprintf.c (struct format_result): Rename member.
6278         (struct fmtresult): Add member and initialize it in ctors.
6279         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
6280         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
6281         (format_directive): Set POSUNDER4K when MAYFAIL is set.
6282         (parse_directive): Handle %C same as %c.
6283         (sprintf_dom_walker::compute_format_length): Adjust.
6284         (is_call_safe): Adjust.
6286 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6288         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
6289         for determining how to count the elements.
6290         * builtins.h (c_strlen): Adjust prototype.
6291         * expr.c (string_constant): Add new parameter mem_size.
6292         Set *mem_size appropriately.
6293         * expr.h (string_constant): Adjust protoype.
6294         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
6295         * gimple-fold.h (get_range_strlen): Adjust prototype.
6296         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
6297         (format_string): Call get_string_length with eltsize.
6299 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
6301         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
6302         to emit the span, rather than setting it as the prefix.
6304 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
6306         * diagnostic-show-locus.c (layout::start_annotation_line): Add
6307         "margin_char" parameter, defaulting to space.  Use it in place
6308         of pp_space for the initial part of the margin.
6309         (layout::print_leading_fixits): Use '+' when filling the margin
6310         of line-insertion fix-it hints.
6312 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6314         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
6315         Delete.
6317 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6319         * config/rs6000/altivec.md: Don't set length attribute to the default
6320         value.
6321         * config/rs6000/darwin.md: Ditto.
6322         * config/rs6000/dfp.md: Ditto.
6323         * config/rs6000/htm.md: Ditto.
6324         * config/rs6000/rs6000.md: Ditto.
6325         * config/rs6000/sync.md: Ditto.
6326         * config/rs6000/vsx.md: Ditto.
6328 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6330         * config/rs6000/altivec.md: Don't set length attribute to the default
6331         value, for branch instructions.
6332         * config/rs6000/darwin.md: Ditto.
6333         * config/rs6000/rs6000.md: Ditto.
6335 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
6337         * config/rs6000/rs6000.md (length): Always define as const_int 4.
6338         (unnamed conditional branch define_insn): Set length to 4 or 8
6339         depending on offset.
6340         (<bd>_<mode>): Similar, for alternative 0.
6341         (<bd>tf_<mode>): Ditto.
6343 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
6345         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
6347 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
6349         * doc/rtl.texi: Replace old RTX class names with new names.
6352 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
6354         * expmed.h (canonicalize_comparison): New declaration.
6355         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
6356         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
6357         * optabs.c (prepare_cmp_insn): Likewise.
6358         * rtl.h (unsigned_condition_p): New function which checks if a
6359         comparison operator is unsigned.
6361 2018-08-16  Nathan Sidwell  <nathan@acm.org>
6363         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
6364         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
6366 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
6368         PR target/84711
6369         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
6370         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
6371         (mov<mov>): ..this and enable unconditionally.
6373 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
6375         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
6377 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
6379         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
6380         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
6381         "Common" with "Target".
6383 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
6385         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
6386         * doc/invoke.texi (mmitigate-rop): Remove.
6387         * config/i386/i386.c: Do not include "regrename.h".
6388         (ix86_rop_should_change_byte_p, reg_encoded_number)
6389         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
6390         Remove.
6391         (ix86_reorg): Remove call to ix86_mitigate_rop.
6392         * config/i386/i386.md (attr "modrm_class"): Remove.
6393         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
6394         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
6395         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
6397 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
6399         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
6400         allow folding of mergeh() and mergel() for the float and double types.
6401         (fold_mergehl_helper): Rework to handle building a permute tree
6402         for float vectors.
6404 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
6406         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
6407         for TARGET_SSE.
6409 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
6411         * common.opt (fdiagnostics-show-labels): New option.
6412         * diagnostic-show-locus.c (class layout_range): Add field
6413         "m_label".
6414         (class layout): Add field "m_show_labels_p".
6415         (layout_range::layout_range): Add param "label" and use it to
6416         initialize m_label.
6417         (make_range): Pass in NULL for new "label" param of layout_range's
6418         ctor.
6419         (layout::layout): Initialize m_show_labels_p.
6420         (layout::maybe_add_location_range): Pass in loc_range->m_label
6421         when constructing layout_range instances.
6422         (struct line_label): New struct.
6423         (layout::print_any_labels): New member function.
6424         (layout::print_line): Call it if label-printing is enabled.
6425         (selftest::test_one_liner_labels): New test.
6426         (selftest::test_diagnostic_show_locus_one_liner): Call it.
6427         * diagnostic.c (diagnostic_initialize): Initialize
6428         context->show_labels_p.
6429         * diagnostic.h (struct diagnostic_context): Add field
6430         "show_labels_p".
6431         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
6432         -fno-diagnostics-show-labels.
6433         * dwarf2out.c (gen_producer_string): Add
6434         OPT_fdiagnostics_show_labels to the ignored options.
6435         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
6436         param.
6437         (gcc_rich_location::maybe_add_expr): Likewise.
6438         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
6439         label" param, defaulting to NULL.
6440         (gcc_rich_location::add_expr): Add "label" param.
6441         (gcc_rich_location::maybe_add_expr): Likewise.
6442         (class text_range_label): New class.
6443         (class range_label_for_type_mismatch): New class.
6444         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
6445         of format_warning_va.
6446         (fmtwarn_n): Likewise for new params of format_warning_n_va.
6447         * lto-wrapper.c (merge_and_complain): Add
6448         OPT_fdiagnostics_show_labels to the "pick one setting" options.
6449         (append_compiler_options): Likewise to the dropped options.
6450         (append_diag_options): Likewise to the passed-on options.
6451         * opts.c (common_handle_option): Handle the new option.
6452         * selftest-diagnostic.c
6453         (test_diagnostic_context::test_diagnostic_context): Enable
6454         show_labels_p.
6455         * substring-locations.c: Include "gcc-rich-location.h".
6456         (format_warning_n_va): Add "fmt_label" and "param_label" params
6457         and use them as appropriate.
6458         (format_warning_va): Add "fmt_label" and "param_label" params,
6459         passing them on to format_warning_n_va.
6460         (format_warning_at_substring): Likewise.
6461         (format_warning_at_substring_n): Likewise.
6462         * substring-locations.h (format_warning_va): Add "fmt_label" and
6463         "param_label" params.
6464         (format_warning_n_va): Likewise.
6465         (format_warning_at_substring): Likewise.
6466         (format_warning_at_substring_n): Likewise.
6467         * toplev.c (general_init): Initialize global_dc->show_labels_p.
6469 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
6471         PR testsuite/86519
6472         * builtins.c (expand_builtin_memcmp): Do not expand the call
6473         when overflow is detected.
6475 2018-08-15  Martin Sebor  <msebor@redhat.com>
6477         PR tree-optimization/71625
6478         * config/aarch64/aarch64-builtins.c
6479         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
6481 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
6483         * config/s390/s390.c (s390_reorg): Remove loop.
6485 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
6487         * config/darwin.c
6488          (darwin_function_switched_text_sections): Delete.
6489         * gcc/config/darwin.h
6490          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
6492 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
6494         PR target/81685
6495         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
6496         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
6497         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
6499 2018-08-15  Martin Liska  <mliska@suse.cz>
6501         PR tree-optimization/86925
6502         * predict.c (expr_expected_value_1): When taking
6503         later predictor, assign also probability.
6504         Use fold_build2_initializer_loc in order to fold
6505         the expression in -frounding-math.
6507 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
6509         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
6510         patterns.
6511         (expand_vec_perm_1): Try the new method.
6513 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
6515         PR target/86547
6516         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
6517         Check whether lra_live_max_point is 0 before dividing.
6519 2018-08-14  Martin Sebor  <msebor@redhat.com>
6521         PR tree-optimization/86650
6522         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
6523         (vrp_prop::check_mem_ref): Same.
6525 2018-08-13  Liu Hao <lh_mouse@126.com>
6527         * pretty-print.c (eat_esc_sequence): Swap the foreground and
6528         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
6529         and clear it thereafter, as it only works for DBCS.
6531 2018-08-13  Liu Hao <lh_mouse@126.com>
6533         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
6534         handle returned by _get_osf_handle().
6536 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
6538         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
6539         for folding vec_perm.
6541 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
6543         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
6544         Add support for gimple-folding of vec_pack() and vec_unpack()
6545         intrinsics.
6547 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
6549         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
6550         vec_xst variants to the list.
6551         (rs6000_gimple_fold_builtin): Add support for folding unaligned
6552         vector loads and stores.
6554 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
6556         * config.gcc (rs6000-ibm-aix4.x): Delete.
6557         (rs6000-ibm-aix5.1): Delete.
6558         (rs6000-ibm-aix5.2): Delete.
6559         (rs6000-ibm-aix5.3): Delete.
6560         * config/rs6000/aix43.h: Delete.
6561         * config/rs6000/aix51.h: Delete.
6562         * config/rs6000/aix52.h: Delete.
6563         * config/rs6000/t-aix43: Delete.
6565 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
6567         * config/s390/s390.c (s390_decompose_constant_pool_ref):
6568         New function.
6569         (s390_decompose_address): Factor out constant pool ref
6570         decomposition.
6572 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6574         * config/nds32/nds32-predicates.c
6575         (nds32_can_use_bclr_p): Change return type as bool.
6576         (nds32_can_use_bset_p): Ditto.
6577         (nds32_can_use_btgl_p): Ditto.
6578         (nds32_can_use_bitci_p): Ditto.
6579         * config/nds32/nds32-protos.h
6580         (nds32_can_use_bclr_p): Change declaration.
6581         (nds32_can_use_bset_p): Ditto.
6582         (nds32_can_use_btgl_p): Ditto.
6583         (nds32_can_use_bitci_p): Ditto.
6585 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6587         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
6588         Support -msched-prolog-epilog option.
6589         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
6591 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6593         * common/config/nds32/nds32-common.c
6594         (nds32_option_optimization_table): Enalbe -malways-align.
6596 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
6598         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
6599         extra_headers.
6600         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
6601         OPT_misr_secure_ case.
6602         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
6603         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
6604         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
6605         secure attribute.
6606         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
6607         (nds32_isr_info): New field security_level.
6608         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
6609         * config/nds32/nds32.md (return_internal): Consider critical attribute.
6610         * config/nds32/nds32.opt (misr-secure): New option.
6611         * config/nds32/nds32_init.inc: New file.
6612         * config/nds32/nds32_isr.h: New file.
6614 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
6616         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
6617         Update comment for atomic instructions.
6618         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
6619         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
6620         Remove.
6621         (atomic_loaddi): Revise fence expansion to only emit fence prior to
6622         load for __ATOMIC_SEQ_CST model.
6623         (atomic_loaddi_1): Remove float register target.
6624         (atomic_storedi): Handle CONST_INT values.
6625         (atomic_storedi_1): Remove float register source.  Add special case
6626         for zero value.
6627         (memory_barrier): New expander and insn.
6629 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
6631         PR tree-optimization/86835
6632         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
6633         new_stmt after def_gsi, make sure to insert new_square_stmt after
6634         that stmt, not 2 stmts before it.
6636 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
6638         PR target/82418
6639         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
6640         instead of SWI48.
6642 2018-08-10  Martin Liska  <mliska@suse.cz>
6644         PR target/83610
6645         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
6646         function type.
6647         * builtins.c (expand_builtin_expect_with_probability):
6648         New function.
6649         (expand_builtin_expect_with_probability): New function.
6650         (build_builtin_expect_predicate): Add new argumnet probability
6651         for BUILT_IN_EXPECT_WITH_PROBABILITY.
6652         (fold_builtin_expect):
6653         (fold_builtin_2):
6654         (fold_builtin_3):
6655         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
6656         * builtins.h (fold_builtin_expect): Set new argument.
6657         * doc/extend.texi: Document __builtin_expect_with_probability.
6658         * doc/invoke.texi: Likewise.
6659         * gimple-fold.c (gimple_fold_call): Pass new argument.
6660         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
6661         also BUILT_IN_EXPECT_WITH_PROBABILITY.
6662         * predict.c (get_predictor_value): New function.
6663         (expr_expected_value): Add new argument probability. Assume
6664         that predictor and probability are always non-null.
6665         (expr_expected_value_1): Likewise.  For __builtin_expect and
6666         __builtin_expect_with_probability set probability.  Handle
6667         combination in binary expressions.
6668         (tree_predict_by_opcode): Simplify code by simply calling
6669         get_predictor_value.
6670         (pass_strip_predict_hints::execute): Add handling of
6671         BUILT_IN_EXPECT_WITH_PROBABILITY.
6672         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
6673         new predictor.
6674         * tree.h (DECL_BUILT_IN_P): New function.
6676 2018-08-10  Martin Liska  <mliska@suse.cz>
6678         PR tree-optimization/85799
6679         * passes.def: Add argument for pass_strip_predict_hints.
6680         * predict.c (class pass_strip_predict_hints): Add new argument
6681         early_p.
6682         (strip_predictor_early): New function.
6683         (pass_strip_predict_hints::execute): Call the function to
6684         strip predictors.
6685         (strip_predict_hints): New function.
6686         * predict.def: Fix comment.
6688 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
6690         * Makefile.in: Clarify which tm.texi to copy over to assert the
6691         right to grant a GFDL license for all.
6693 2018-08-09  Jeff Law  <law@redhat.com>
6695         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
6696         unused variable.
6698 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
6700         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
6701         prototype.
6703 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6705         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
6706         reductions for variable-length vectors.
6708 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
6710         PR other/84889
6711         * common.opt (fdiagnostics-show-line-numbers): New option.
6712         * diagnostic-show-locus.c (class layout): Add fields
6713         "m_show_line_numbers_p" and "m_linenum_width";
6714         (num_digits): New function.
6715         (test_num_digits): New function.
6716         (layout::layout): Initialize new fields.  Update m_x_offset
6717         logic to handle any left margin.
6718         (layout::print_source_line): Print line number when requested.
6719         (layout::start_annotation_line): New member function.
6720         (layout::print_annotation_line): Call it.
6721         (layout::print_leading_fixits): Likewise.
6722         (layout::print_trailing_fixits): Likewise.  Update calls to
6723         move_to_column for new parameter.
6724         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
6725         it to potentially call start_annotation_line.
6726         (layout::show_ruler): Call start_annotation_line.
6727         (selftest::test_line_numbers_multiline_range): New selftest.
6728         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
6729         and selftest::test_line_numbers_multiline_range.
6730         * diagnostic.c (diagnostic_initialize): Initialize
6731         show_line_numbers_p.
6732         * diagnostic.h (struct diagnostic_context): Add field
6733         "show_line_numbers_p".
6734         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
6735         -fno-diagnostics-show-line-numbers.
6736         * dwarf2out.c (gen_producer_string): Add
6737         OPT_fdiagnostics_show_line_numbers to the ignored options.
6738         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
6739         one setting" options.
6740         (append_compiler_options): Likewise to the dropped options.
6741         (append_diag_options): Likewise to the passed-on options.
6742         * opts.c (common_handle_option): Handle the new option.
6743         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
6745 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6747         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
6748         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
6749         third argument of type "const signed char" to descriptions of
6750         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
6751         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
6752         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
6753         __builtin_bcdsub_ov functions.
6755 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6757         PR tree-optimization/86858
6758         * tree-vect-loop.c (vect_is_simple_reduction): Restore
6759         flow_bb_inside_loop_p calls.
6761 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
6763         PR tree-optimization/86871
6764         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
6765         instead of gimple_assign_lhs.
6767 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
6769         PR target/86887
6770         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
6771         register constraint to operand 0.
6772         (add<mode>3_carryinC): Likewise.
6773         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
6775 2018-08-09  Martin Liska  <mliska@suse.cz>
6777         PR c/86895
6778         * common.opt: Remove extra line.
6780 2018-08-09  Martin Liska  <mliska@suse.cz>
6782         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
6783         at the end of a line, make first letter capital and end up
6784         a sentence with a dot.
6785         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
6786         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
6787         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
6788         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
6789         (PARAM_MAX_ISL_OPERATIONS): Likewise.
6790         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
6791         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
6792         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
6793         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
6794         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
6795         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
6796         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
6797         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
6798         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
6799         (PARAM_TREE_REASSOC_WIDTH): Likewise.
6800         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
6801         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
6802         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
6804 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
6806         PR target/84332
6807         * config/s390/s390.c (s390_option_override_internal): Reduce the
6808         stack-clash-protection-probe-interval param if it would be too big
6809         for z900.
6811 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
6813         PR target/46179
6814         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
6815         * config/m68k/m68k.c (handle_move_double): Don't call
6816         m68k_final_prescan_insn.
6817         (m68k_adjust_decorated_operand): Renamed from
6818         m68k_final_prescan_insn, remove first and third operand and
6819         simplify.
6820         (print_operand): Call it.
6821         (print_operand_address): Call it.
6823 2018-08-08  Nathan Sidwell  <nathan@acm.org>
6825         * diagnostic.c (diagnostic_report_current_module): Use
6826         linemap_included_from & linemap_included_from_linemap.
6828 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
6830         * config/aarch64/aarch64-cores.def: Add phecda core.
6831         * config/aarch64/aarch64-tune.md: Regenerate.
6832         * doc/invoke.texi: Add phecda core.
6834 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
6836         PR target/85295
6837         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
6838         definitions.
6839         * config/s390/s390.md ("movti"): Add more alternatives for
6840         constant to GPR copies.
6842 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
6844         * config/s390/s390.c: Fix whitespace damage throughout the file.
6845         * config/s390/s390.h: Likewise.
6846         * config/s390/tpf.h: Likewise.
6848 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
6850         * config/s390/s390.c (s390_loadrelative_operand_p):
6851         Remove TARGET_CPU_ZARCH usages.
6852         (s390_rtx_costs): Likewise.
6853         (s390_legitimate_constant_p): Likewise.
6854         (s390_cannot_force_const_mem): Likewise.
6855         (legitimate_reload_constant_p): Likewise.
6856         (s390_preferred_reload_class): Likewise.
6857         (legitimize_pic_address): Likewise.
6858         (legitimize_tls_address): Likewise.
6859         (s390_split_branches): Removed.
6860         (s390_add_execute): Removed.
6861         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
6862         (s390_mainpool_start): Likewise.
6863         (s390_mainpool_finish): Likewise.
6864         (s390_mainpool_cancel): Removed.
6865         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
6866         (s390_chunkify_cancel): Likewise.
6867         (s390_return_addr_rtx): Likewise.
6868         (s390_register_info): Remove split_branches_pending_p uages.
6869         (s390_optimize_register_info): Likewise.
6870         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
6871         split_branches_pending_p usages.
6872         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
6873         (s390_load_got): Likewise.
6874         (s390_expand_split_stack_prologue): Likewise.
6875         (output_asm_nops): Likewise.
6876         (s390_function_profiler): Likewise.
6877         (s390_emit_call): Likewise.
6878         (s390_conditional_register_usage): Likewise.
6879         (s390_optimize_prologue): Likewise.
6880         (s390_reorg): Remove TARGET_CPU_ZARCH and
6881         split_branches_pending_p usages.
6882         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
6883         usages.
6884         (s390_output_indirect_thunk_function): Likewise.
6885         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
6886         (TARGET_CPU_ZARCH_P): Removed.
6887         (struct machine_function): Remove split_branches_pending_p.
6888         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
6890 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
6892         * common/config/s390/s390-common.c (processor_flags_table):
6893         Remove flags.
6894         * config.gcc: Remove with_arch/with_tune support.
6895         * config/s390/2064.md: Remove cpu attribute comparisons.
6896         * config/s390/driver-native.c (s390_host_detect_local_cpu):
6897         Remove MTN.
6898         * config/s390/linux.h (ASM_SPEC):
6899         Remove -march support.
6900         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
6901         Use a table to get an arch level.
6902         * config/s390/s390-opts.h (enum processor_type):
6903         Remove enum values.
6904         * config/s390/s390.c
6905         (processor_table): Remove entries, add arch_level values.
6906         (s390_issue_rate): Remove cases.
6907         (s390_option_override): Adjust
6908         s390_option_override_internal() call.
6909         (s390_option_override_internal): Remove deprecation warning.
6910         (s390_valid_target_attribute_tree): Adjust
6911         s390_option_override_internal() call.
6912         * config/s390/s390.h (struct s390_processor):
6913         Share with s390-c.c, add arch_level field.
6914         * config/s390/s390.md:
6915         Remove occurrences in cpu attribute.
6916         * config/s390/s390.opt: Remove -march/-mtune support.
6917         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
6918         * doc/invoke.texi: Remove deprecation warning.
6920 2018-08-08  Luis Machado  <luis.machado@linaro.org>
6922         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
6923         global.
6924         (qdf24xx_tunings): Set vector cost structure to
6925         qdf24xx_vector_cost.
6927         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
6928         <register_sextend>: Set to 3.
6930 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
6932         PR target/86838
6933         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
6934         * config/aarch64/aarch64-simd.md
6935         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
6936         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
6937         (aarch64_frecpx<mode>): ...this new pattern.
6938         * config/aarch64/aarch64-simd-builtins.def: Remove comment
6939         about aarch64_frecp<FRECP:frecp_suffix><mode>.
6941 2018-08-07  Martin Liska  <mliska@suse.cz>
6943         PR middle-end/83023
6944         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
6945         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
6946         * predict.def (PRED_MALLOC_NONNULL): New predictor.
6947         * doc/extend.texi: Document that malloc attribute adds
6948         hit to compiler.
6950 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
6952         PR target/86785
6953         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6954         Define to speculation_safe_value_not_needed.
6956 2018-08-06  Jeff Law  <law@redhat.com>
6958         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
6959         the vr_values instance to cprop_into_stmt.
6960         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
6961         (cprop_operand): Also query EVRP to determine if OP is a constant.
6963 2018-08-06  Nathan Sidwell  <nathan@acm.org>
6965         * diagnostic.c (diagnostic_report_current_module): Reroll
6966         included-at loop.  Translate text.
6968 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
6970         * function-tests.c (selftest::test_expansion_to_rtl): Call
6971         free_after_compilation.
6973 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
6975         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
6977 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
6979         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
6980         loops with memory block operations from getting unrolled.
6982 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
6984         PR target/86807
6985         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6986         Define to speculation_safe_value_not_needed.
6988 2018-08-06  Jeff Law  <law@redhat.com>
6990         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
6991         assert.
6993 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
6995         PR target/86662
6996         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
6997         with all enabled __intN types.
6999         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
7001 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7003         * alias.c (record_set): Check for clobber high.
7004         * cfgexpand.c (expand_gimple_stmt): Likewise.
7005         * combine-stack-adj.c (single_set_for_csa): Likewise.
7006         * combine.c (find_single_use_1): Likewise.
7007         (set_nonzero_bits_and_sign_copies): Likewise.
7008         (get_combine_src_dest): Likewise.
7009         (is_parallel_of_n_reg_sets): Likewise.
7010         (try_combine): Likewise.
7011         (record_dead_and_set_regs_1): Likewise.
7012         (reg_dead_at_p_1): Likewise.
7013         (reg_dead_at_p): Likewise.
7014         * dce.c (deletable_insn_p): Likewise.
7015         (mark_nonreg_stores_1): Likewise.
7016         (mark_nonreg_stores_2): Likewise.
7017         * df-scan.c (df_find_hard_reg_defs): Likewise.
7018         (df_uses_record): Likewise.
7019         (df_get_call_refs): Likewise.
7020         * dwarf2out.c (mem_loc_descriptor): Likewise.
7021         * haifa-sched.c (haifa_classify_rtx): Likewise.
7022         * ira-build.c (create_insn_allocnos): Likewise.
7023         * ira-costs.c (scan_one_insn): Likewise.
7024         * ira.c (equiv_init_movable_p): Likewise.
7025         (rtx_moveable_p): Likewise.
7026         (interesting_dest_for_shprep): Likewise.
7027         * jump.c (mark_jump_label_1): Likewise.
7028         * postreload-gcse.c (record_opr_changes): Likewise.
7029         * postreload.c (reload_cse_simplify): Likewise.
7030         (struct reg_use): Add source expr.
7031         (reload_combine): Check for clobber high.
7032         (reload_combine_note_use): Likewise.
7033         (reload_cse_move2add): Likewise.
7034         (move2add_note_store): Likewise.
7035         * print-rtl.c (print_pattern): Likewise.
7036         * recog.c (decode_asm_operands): Likewise.
7037         (store_data_bypass_p): Likewise.
7038         (if_test_bypass_p): Likewise.
7039         * regcprop.c (kill_clobbered_value): Likewise.
7040         (kill_set_value): Likewise.
7041         * reginfo.c (reg_scan_mark_refs): Likewise.
7042         * reload1.c (maybe_fix_stack_asms): Likewise.
7043         (eliminate_regs_1): Likewise.
7044         (elimination_effects): Likewise.
7045         (mark_not_eliminable): Likewise.
7046         (scan_paradoxical_subregs): Likewise.
7047         (forget_old_reloads_1): Likewise.
7048         * reorg.c (find_end_label): Likewise.
7049         (try_merge_delay_insns): Likewise.
7050         (redundant_insn): Likewise.
7051         (own_thread_p): Likewise.
7052         (fill_simple_delay_slots): Likewise.
7053         (fill_slots_from_thread): Likewise.
7054         (dbr_schedule): Likewise.
7055         * resource.c (update_live_status): Likewise.
7056         (mark_referenced_resources): Likewise.
7057         (mark_set_resources): Likewise.
7058         * rtl.c (copy_rtx): Likewise.
7059         * rtlanal.c (reg_referenced_p): Likewise.
7060         (single_set_2): Likewise.
7061         (noop_move_p): Likewise.
7062         (note_stores): Likewise.
7063         * sched-deps.c (sched_analyze_reg): Likewise.
7064         (sched_analyze_insn): Likewise.
7066 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7068         * cse.c (invalidate_reg): New function extracted from...
7069         (invalidate): ...here.
7070         (canonicalize_insn): Check for clobber high.
7071         (invalidate_from_clobbers): invalidate clobber highs.
7072         (invalidate_from_sets_and_clobbers): Likewise.
7073         (count_reg_usage): Check for clobber high.
7074         (insn_live_p): Likewise.
7075         * cselib.c (cselib_expand_value_rtx_1):Likewise.
7076         (cselib_invalidate_regno): Check for clobber in setter.
7077         (cselib_invalidate_rtx): Pass through setter.
7078         (cselib_invalidate_rtx_note_stores):
7079         (cselib_process_insn): Check for clobber high.
7080         * cselib.h (cselib_invalidate_rtx): Add operand.
7082 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7084         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
7085         (mark_not_eliminable): Likewise.
7086         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
7087         * lra-lives.c (process_bb_lives): Check for clobber high.
7088         * lra.c (new_insn_reg): Remember clobber highs.
7089         (collect_non_operand_hard_regs): Check for clobber high.
7090         (lra_set_insn_recog_data): Likewise.
7091         (add_regs_to_insn_regno_info): Likewise.
7092         (lra_update_insn_regno_info): Likewise.
7094 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7096         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
7097         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
7099 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7101         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
7102         (copy_insn_1): Likewise.
7103         (gen_hard_reg_clobber_high): New gen function.
7104         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
7105         * genemit.c (gen_exp): Likewise.
7106         (gen_emit_seq): Pass through info.
7107         (gen_insn): Check for CLOBBER_HIGH.
7108         (gen_expand): Pass through info.
7109         (gen_split): Likewise.
7110         (output_add_clobbers): Likewise.
7111         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
7112         (remove_clobbers): Likewise.
7113         * rtl.h (gen_hard_reg_clobber_high): New declaration.
7115 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
7117         * doc/rtl.texi (clobber_high): Add.
7118         (parallel): Add in clobber high
7119         * rtl.c (rtl_check_failed_code3): Add function.
7120         * rtl.def (CLOBBER_HIGH): Add expression.
7121         * rtl.h (RTL_CHECKC3): Add macro.
7122         (rtl_check_failed_code3): Add declaration.
7123         (XC3EXP): Add macro.
7125 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
7127         PR target/86386
7128         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
7129         cfun->machine->max_used_stack_alignment if needed.
7131 2018-08-04  Martin Sebor  <msebor@redhat.com>
7133         PR tree-optimization/86571
7134         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
7135         NaN output to 4.
7137 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
7139         PR target/86799
7140         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7141         Define.
7143 2018-08-03  Jeff Law  <law@redhat.com>
7145         PR target/86795
7146         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7147         Define to speculation_safe_value_not_needed.
7149 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
7151         * doc/gcov.texi (-x): Remove duplicate "to".
7152         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
7153         (-Wif-not-aligned): Remove duplicate "is".
7154         (-flto): Remove duplicate "the".
7155         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
7156         duplicate "v5.00.b".
7157         (MSP430 Options): Remove duplicate "and" from the description
7158         of "-mgprel-sec=regexp".
7159         (x86 Options): Remove duplicate copies of "vmldLog102" and
7160         vmlsLog104 from description of "-mveclibabi=type".
7162 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
7164         * internal-fn.h (first_commutative_argument): Declare.
7165         * internal-fn.c (first_commutative_argument): New function.
7166         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
7167         restrictions for pattern statements.  Use first_commutative_argument
7168         to look for commutative operands in calls to internal functions.
7170 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
7172         * Makefile.in (wide-int-range.o): New.
7173         * tree-vrp.c: Move all the wide_int_* functions to...
7174         * wide-int-range.cc: ...here.
7175         * tree-vrp.h: Move all the wide_int_* prototypes to...
7176         * wide-int-range.h: ...here.
7178 2018-08-03  Tom de Vries  <tdevries@suse.de>
7180         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
7181         UI_NONE.
7182         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
7183         * except.c (output_function_exception_table): Do early exit if
7184         targetm_common.except_unwind_info (&global_options) == UI_NONE.
7186 2018-08-03  Martin Liska  <mliska@suse.cz>
7188         * predict.c (dump_prediction): Change to 2 digits
7189         in fraction part.
7191 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
7193         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
7194         neon_dup_q to...
7195         (falkor_am_1_gtov_gtov): ... a new insn reservation.
7197 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
7199         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
7200         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
7201         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
7202         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
7203         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
7205 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
7207         * diagnostic-show-locus.c (diagnostic_show_locus): Use
7208         pp_take_prefix when saving the existing prefix.
7209         * diagnostic.c (diagnostic_append_note): Likewise.
7210         * langhooks.c (lhd_print_error_function): Likewise.
7211         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
7212         param's type.  Free the existing prefix.
7213         (pp_take_prefix): New function.
7214         (pretty_printer::pretty_printer): Drop the prefix parameter.
7215         Rename the length parameter to match the comment.
7216         (pretty_printer::~pretty_printer): Free the prefix.
7217         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
7218         parameter.
7219         (struct pretty_printer): Drop the "const" from "prefix" field's
7220         type and clarify memory management.
7221         (pp_set_prefix): Drop the "const" from the 2nd param.
7222         (pp_take_prefix): New decl.
7224 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
7226         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
7227         for word_mode_ok here instead of passing as argument.
7228         (expand_block_compare): Change select_block_compare_mode() call.
7229         (expand_strncmp_gpr_sequence): New function.
7230         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
7232 2018-08-02  Jeff Law  <law@redhat.com>
7234         PR target/86790
7235         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7236         Define to speculation_safe_value_not_needed.
7238         PR target/86784
7239         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7240         Define to speculation_safe_value_not_needed.
7242 2018-08-02  Tom de Vries  <tdevries@suse.de>
7244         PR target/86660
7245         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
7246         function.  Return UI_TARGET unconditionally.
7247         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
7248         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
7250 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
7252         * genemit.c (print_overload_test): Fix typo.
7254 2018-08-02  Richard Biener  <rguenther@suse.de>
7256         PR tree-optimization/86816
7257         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
7258         which checks for value availability before querying it.
7259         (gvn_uses_equal): Use it.
7260         (same_succ_hash): Likewise.
7261         (gimple_equal_p): Likewise.
7263 2018-08-02  Nick Clifton  <nickc@redhat.com>
7265         PR target/86813
7266         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7267         Define to speculation_safe_value_not_needed.
7269         PR target/86810
7270         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7271         Define to speculation_safe_value_not_needed.
7273         PR target/86810
7274         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7275         Define to speculation_safe_value_not_needed.
7277         PR target/86803
7278         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7279         Define to speculation_safe_value_not_needed.
7281         PR target/86797
7282         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7283         Define to speculation_safe_value_not_needed.
7285         PR target/86791
7286         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7287         Define to speculation_safe_value_not_needed.
7289         PR target/86789
7290         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7291         Define to speculation_safe_value_not_needed.
7293         PR target/86787
7294         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7295         Define to speculation_safe_value_not_needed.
7297         PR target/86782
7298         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
7299         speculation_safe_value_not_needed.
7301         PR target/86781
7302         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
7303         to speculation_safe_value_not_needed.
7305 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
7307         * doc/md.texi: Expand the documentation of instruction names
7308         to mention port-local uses.  Document '@' in pattern names.
7309         * read-md.h (overloaded_instance, overloaded_name): New structs.
7310         (mapping): Declare.
7311         (md_reader::handle_overloaded_name): New member function.
7312         (md_reader::get_overloads): Likewise.
7313         (md_reader::m_first_overload): New member variable.
7314         (md_reader::m_next_overload_ptr): Likewise.
7315         (md_reader::m_overloads_htab): Likewise.
7316         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
7317         m_next_overload_ptr and m_overloads_htab.
7318         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
7319         (get_mode_token, get_code_token, get_int_token): New functions.
7320         (map_attr_string): Add an optional argument that passes back
7321         the associated iterator.
7322         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
7323         (md_reader::handle_overloaded_name, add_overload_instance): New
7324         functions.
7325         (apply_iterators): Handle '@' names.  Report an error if '@'
7326         is used without iterators.
7327         (initialize_iterators): Initialize the new iterator_group fields.
7328         * genopinit.c (handle_overloaded_code_for)
7329         (handle_overloaded_gen): New functions.
7330         (main): Use them to print declarations of maybe_code_for_* and
7331         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
7332         * genemit.c (print_overload_arguments, print_overload_test)
7333         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
7334         (main): Use it to print definitions of maybe_code_for_* and
7335         maybe_gen_* functions.
7336         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
7337         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
7338         instead of explicit mode checks.
7339         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
7340         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
7341         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
7342         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
7343         (aarch64_expand_compare_and_swap): Likewise
7344         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
7345         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
7346         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
7347         (aarch64_constant_pool_reload_icode): Delete.
7348         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
7349         instead of aarch64_constant_pool_reload_icode.  Use
7350         code_for_aarch64_reload_mov instead of explicit mode checks.
7351         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
7352         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
7353         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
7354         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
7355         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
7356         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
7357         (aarch64_atomic_load_op_code): Delete.
7358         (aarch64_emit_atomic_load_op): Likewise.
7359         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
7360         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
7361         instead of aarch64_emit_atomic_load_op.
7362         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
7363         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
7364         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
7365         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
7366         character before the pattern name.
7367         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
7368         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
7369         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
7370         (aarch64_frecps<mode>): Likewise.
7371         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
7372         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
7373         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
7374         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
7375         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
7377 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
7379         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
7380         Allow HFmode constants if TARGET_FP_F16INST.
7382 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
7384         PR target/86014
7385         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
7386         No longer check last store for clobber of address register.
7388 2018-08-02  Martin Liska  <mliska@suse.cz>
7390         PR gcov-profile/86817
7391         * gcov.c (process_all_functions): New function.
7392         (main): Call it.
7393         (process_file): Move functions processing to
7394         process_all_functions.
7396 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
7398         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
7399         "const" to the "gimple *" and "rtx_insn *" parameters.
7400         * dumpfile.h (dump_user_location_t::dump_user_location_t):
7401         Likewise.
7402         (dump_location_t::dump_location_t): Likewise.
7404 2018-08-01  Martin Sebor  <msebor@redhat.com>
7406         PR tree-optimization/86650
7407         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
7408         rather than a "gcall *".  Directly pass the data of interest
7409         to percent_K_format, rather than building a temporary CALL_EXPR
7410         to hold it.
7411         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
7412         (gimple_fold_builtin_strncat): Adjust.
7413         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
7414         gcall* argument with gimple*.
7415         * gimple-ssa-warn-restrict.c (check_call): Same.
7416         (wrestrict_dom_walker::before_dom_children): Same.
7417         (builtin_access::builtin_access): Same.
7418         (check_bounds_or_overlap): Same
7419         (maybe_diag_overlap): Same.
7420         (maybe_diag_offset_bounds): Same.
7421         * tree-diagnostic.c (default_tree_printer): Move usage of
7422         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
7423         to this callsite.
7424         * tree-pretty-print.c (percent_K_format): Add argument.
7425         * tree-pretty-print.h: Add argument.
7426         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
7427         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
7428         (maybe_diag_stxncpy_trunc): Same.
7429         (handle_builtin_stxncpy): Same.
7430         (handle_builtin_strcat): Same.
7432 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7434         * match.pd: Optimise pointer range checks.
7436 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7438         PR tree-optimization/86758
7439         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
7440         to remove pattern statements.
7442 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7444         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
7445         result of dfs_enumerate_from when constructing stmt_vec_infos,
7446         instead of additionally calling get_loop_body.
7448 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7450         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
7451         parameter.
7452         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
7453         When creating an iv, assert that the step is not known to be zero.
7454         (vect_setup_realignment): Update call accordingly.
7455         * tree-vect-stmts.c (vectorizable_store): Likewise.
7456         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
7458 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7460         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
7461         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
7462         (vectorizable_reduction): Likewise.
7463         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7464         (vect_detect_hybrid_slp_stmts): Likewise.
7465         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7467 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
7469         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
7470         (wide_int_set_zero_nonzero_bits): ...this.
7471         (zero_nonzero_bits_from_vr): Rename to...
7472         (vrp_set_zero_nonzero_bits): ...this.
7473         (extract_range_from_multiplicative_op_1): Abstract wide int
7474         code...
7475         (wide_int_range_multiplicative_op): ...here.
7476         (extract_range_from_binary_expr_1): Extract wide int binary
7477         operations into their own functions.
7478         (wide_int_range_lshift): New.
7479         (wide_int_range_can_optimize_bit_op): New.
7480         (wide_int_range_shift_undefined_p): New.
7481         (wide_int_range_bit_xor): New.
7482         (wide_int_range_bit_ior): New.
7483         (wide_int_range_bit_and): New.
7484         (wide_int_range_trunc_mod): New.
7485         (extract_range_into_wide_ints): New.
7486         (vrp_shift_undefined_p): New.
7487         (extract_range_from_multiplicative_op): New.
7488         (vrp_can_optimize_bit_op): New.
7489         * tree-vrp.h (value_range::dump): New.
7490         (wide_int_range_multiplicative_op): New.
7491         (wide_int_range_lshift):New.
7492         (wide_int_range_shift_undefined_p): New.
7493         (wide_int_range_bit_xor): New.
7494         (wide_int_range_bit_ior): New.
7495         (wide_int_range_bit_and): New.
7496         (wide_int_range_trunc_mod): New.
7497         (zero_nonzero_bits_from_bounds): Rename to...
7498         (wide_int_set_zero_nonzero_bits): ...this.
7499         (zero_nonzero_bits_from_vr): Rename to...
7500         (vrp_set_zero_nonzero_bits): ...this.
7501         (range_easy_mask_min_max): Rename to...
7502         (wide_int_range_can_optimize_bit_op): this.
7503         * vr-values.c (simplify_bit_ops_using_ranges): Rename
7504         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
7506 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7508         * tree-vectorizer.h (vect_orig_stmt): New function.
7509         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
7510         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
7511         (vect_create_epilog_for_reduction): Likewise.
7512         (vectorizable_live_operation): Likewise.
7513         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
7514         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
7515         * tree-vect-stmts.c (vectorizable_call): Likewise.
7516         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
7518 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7520         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
7521         argument.
7522         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
7523         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
7524         (vect_transform_loop): Likewise.
7525         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
7527 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7529         * tree-vectorizer.h (vect_schedule_slp): Return void.
7530         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
7531         (vect_schedule_slp): Likewise.
7533 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7535         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
7536         argument.
7537         (vect_transform_loop): Update calls accordingly.  Schedule SLP
7538         instances before the main loop, if any exist.
7540 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
7542         PR tree-optimization/86749
7543         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
7544         If the lhs is used in a COND_EXPR, check that it is being used
7545         as the "then" or "else" value.
7547 2018-08-01  Tom de Vries  <tdevries@suse.de>
7549         PR target/86800
7550         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
7551         speculation_safe_value_not_needed.
7553 2018-08-01  Richard Biener  <rguenther@suse.de>
7555         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
7556         as base and offset.
7558 2018-08-01  Martin Liska  <mliska@suse.cz>
7560         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
7561         format how successful transformation is dumped.
7562         (gimple_mod_pow2_value_transform): Likewise.
7563         (gimple_mod_subtract_transform): Likewise.
7564         (gimple_stringops_transform): Likewise.
7566 2018-08-01  Martin Liska  <mliska@suse.cz>
7568         PR value-prof/35543
7569         * value-prof.c (interesting_stringop_to_profile_p):
7570         Simplify the code and add BUILT_IN_MEMMOVE.
7571         (gimple_stringops_transform): Likewise.
7573 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
7575         * config/aarch64/aarch64-simd.md
7576         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
7577         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
7578         use GPI iterator instead of SI mode.
7580 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
7582         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
7583         rs6000_speculation_barrier.
7584         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
7585         new barrier pattern name.
7587 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
7589         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
7590         (speculation_barrier): New insn.
7592 2018-08-01  Richard Biener  <rguenther@suse.de>
7594         PR bootstrap/86724
7595         * graphite.h: Include isl/id.h and isl/space.h to allow build
7596         with ISL 0.20.
7598 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
7600         PR target/86651
7601         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7602         mode for COFF targets.
7603         * defaults.h (TARGET_COFF): Define.
7604         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
7605         TARGET_COFF): Define.
7606         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
7607         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
7608         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
7610 2018-07-31  Alexandre Oliva <oliva@adacore.com>
7612         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
7613         * gimple-streamer-out.c (output_bb): Save it.
7614         * lto-streamer-in.c (input_struct_function_base): Restore
7615         instance discriminator if available.  Create map on demand.
7616         * lto-streamer-out.c (output_struct_function_base): Save it if
7617         available.
7618         * final.c (decl_to_instance_map): Document LTO strategy.
7620 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
7621             Olivier Hainque  <hainque@adacore.com>
7623         * debug.h (decl_to_instance_map_t): New type.
7624         (decl_to_instance_map): Declare.
7625         (maybe_create_decl_to_instance_map): New inline function.
7626         * final.c (bb_discriminator, last_bb_discriminator): New statics,
7627         to track basic block discriminators.
7628         (final_start_function_1): Initialize them.
7629         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
7630         bb_discriminator.
7631         (decl_to_instance_map): New variable.
7632         (map_decl_to_instance, maybe_set_discriminator): New functions.
7633         (notice_source_line): Set discriminator.
7635 2018-07-31  Ian Lance Taylor  <iant@golang.org>
7637         * targhooks.c (default_have_speculation_safe_value): Add
7638         ATTRIBUTE_UNUSED.
7640 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
7642         * dump-context.h: Include "pretty-print.h".
7643         (dump_context::refresh_dumps_are_enabled): New decl.
7644         (dump_context::emit_item): New decl.
7645         (class dump_context): Add fields "m_test_pp" and
7646         "m_test_pp_flags".
7647         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
7648         (temp_dump_context::get_dumped_text): New decl.
7649         (class temp_dump_context): Add field "m_pp".
7650         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
7651         (dump_context::refresh_dumps_are_enabled): ...and add a test for
7652         m_test_pp.
7653         (set_dump_file): Update for above change.
7654         (set_alt_dump_file): Likewise.
7655         (dump_loc): New overload, taking a pretty_printer *.
7656         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
7657         to any test pretty-printer.
7658         (make_item_for_dump_gimple_stmt): New function, adapted from
7659         optinfo::add_gimple_stmt.
7660         (dump_context::dump_gimple_stmt): Call it, and use the result,
7661         eliminating the direct usage of dump_file and alt_dump_file in
7662         favor of indirectly using them via emit_item.
7663         (make_item_for_dump_gimple_expr): New function, adapted from
7664         optinfo::add_gimple_expr.
7665         (dump_context::dump_gimple_expr): Call it, and use the result,
7666         eliminating the direct usage of dump_file and alt_dump_file in
7667         favor of indirectly using them via emit_item.
7668         (make_item_for_dump_generic_expr): New function, adapted from
7669         optinfo::add_tree.
7670         (dump_context::dump_generic_expr): Call it, and use the result,
7671         eliminating the direct usage of dump_file and alt_dump_file in
7672         favor of indirectly using them via emit_item.
7673         (make_item_for_dump_printf_va): New function, adapted from
7674         optinfo::add_printf_va.
7675         (make_item_for_dump_printf): New function.
7676         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
7677         and use the result, eliminating the direct usage of dump_file and
7678         alt_dump_file in favor of indirectly using them via emit_item.
7679         (make_item_for_dump_dec): New function.
7680         (dump_context::dump_dec): Call it, and use the result,
7681         eliminating the direct usage of dump_file and alt_dump_file in
7682         favor of indirectly using them via emit_item.
7683         (make_item_for_dump_symtab_node): New function, adapted from
7684         optinfo::add_symtab_node.
7685         (dump_context::dump_symtab_node): Call it, and use the result,
7686         eliminating the direct usage of dump_file and alt_dump_file in
7687         favor of indirectly using them via emit_item.
7688         (dump_context::begin_scope): Reimplement, avoiding direct usage
7689         of dump_file and alt_dump_file in favor of indirectly using them
7690         via emit_item.
7691         (dump_context::emit_item): New member function.
7692         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
7693         Set up test pretty-printer on the underlying context.  Call
7694         refresh_dumps_are_enabled.
7695         (temp_dump_context::~temp_dump_context): Call
7696         refresh_dumps_are_enabled.
7697         (temp_dump_context::get_dumped_text): New member function.
7698         (selftest::verify_dumped_text): New function.
7699         (ASSERT_DUMPED_TEXT_EQ): New macro.
7700         (selftest::test_capture_of_dump_calls): Run all tests twice, with
7701         and then without optinfo enabled.  Add uses of
7702         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
7703         * dumpfile.h: Update comment for the dump_* API.
7704         * optinfo-emit-json.cc
7705         (selftest::test_building_json_from_dump_calls): Update for new
7706         param for temp_dump_context ctor.
7707         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
7708         and "m_owned" field.
7709         (optinfo_item::~optinfo_item): Likewise.
7710         (optinfo::add_item): New member function.
7711         (optinfo::emit): Update comment.
7712         (optinfo::add_string): Delete.
7713         (optinfo::add_printf): Delete.
7714         (optinfo::add_printf_va): Delete.
7715         (optinfo::add_gimple_stmt): Delete.
7716         (optinfo::add_gimple_expr): Delete.
7717         (optinfo::add_tree): Delete.
7718         (optinfo::add_symtab_node): Delete.
7719         (optinfo::add_dec): Delete.
7720         * optinfo.h (class dump_context): New forward decl.
7721         (optinfo::add_item): New decl.
7722         (optinfo::add_string): Delete.
7723         (optinfo::add_printf): Delete.
7724         (optinfo::add_printf_va): Delete.
7725         (optinfo::add_gimple_stmt): Delete.
7726         (optinfo::add_gimple_expr): Delete.
7727         (optinfo::add_tree): Delete.
7728         (optinfo::add_symtab_node): Delete.
7729         (optinfo::add_dec): Delete.
7730         (optinfo::add_poly_int): Delete.
7731         (optinfo_item::optinfo_item): Remove "owned" param.
7732         (class optinfo_item): Remove field "m_owned".
7734 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
7736         PR middle-end/86705
7737         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
7738         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
7739         requested variable alignment.
7740         (expand_one_ssa_partition): Likewise.
7741         (expand_one_var): Likewise.
7743 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7745         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7746         to speculation_safe_value_not_needed.
7748 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7750         * targhooks.h (speculation_safe_value_not_needed): New prototype.
7751         * targhooks.c (speculation_safe_value_not_needed): New function.
7752         * target.def (have_speculation_safe_value): Update documentation.
7753         * doc/tm.texi: Regenerated.
7755 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7757         * config/aarch64/iterators.md (ALLI_TI): New iterator.
7758         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
7759         expand.
7760         (despeculate_copy<ALLI:mode>_insn): New insn.
7761         (despeculate_copyti_insn): New insn.
7762         (despeculate_simple<ALLI:mode>): New insn
7763         (despeculate_simpleti): New insn.
7764         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
7765         function.
7766         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
7767         aarch64_speculation_safe_value.
7768         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
7770 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7772         * config/aarch64/aarch64-speculation.cc: New file.
7773         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
7774         before pass_reorder_blocks.
7775         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
7776         prototype.
7777         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
7778         X14 and X15 when tracking speculation.
7779         * config/aarch64/aarch64.md (register name constants): Add
7780         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
7781         (unspec): Add UNSPEC_SPECULATION_TRACKER.
7782         (speculation_barrier): New insn attribute.
7783         (cmp<mode>): Allow SP in comparisons.
7784         (speculation_tracker): New insn.
7785         (speculation_barrier): Add speculation_barrier attribute.
7786         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
7787         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
7788         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
7790 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7792         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
7793         aarch64_track_speculation is true.
7794         (tb<optab><mode>1): Likewise.
7795         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
7796         generate CB[N]Z when tracking speculation.
7797         (aarch64_split_compare_and_swap): Likewise.
7798         (aarch64_split_atomic_op): Likewise.
7800 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7802         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
7804 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7806         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
7807         (speculation_barrier): New insn.
7809 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7811         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
7812         * config/arm/arm.md (speculation_barrier): New expand.
7813         (speculation_barrier_insn): New pattern.
7815 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
7817         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
7818         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
7819         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
7820         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
7821         list.
7822         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
7823         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
7824         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
7825         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
7826         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
7827         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
7828         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
7829         * builtins.c (expand_speculation_safe_value): New function.
7830         (expand_builtin): Call it.
7831         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
7832         * doc/extend.texi: Document __builtin_speculation_safe_value.
7833         * doc/md.texi: Document "speculation_barrier" pattern.
7834         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
7835         TARGET_HAVE_SPECULATION_SAFE_VALUE.
7836         * doc/tm.texi: Regenerated.
7837         * target.def (have_speculation_safe_value, speculation_safe_value): New
7838         hooks.
7839         * targhooks.c (default_have_speculation_safe_value): New function.
7840         (default_speculation_safe_value): New function.
7841         * targhooks.h (default_have_speculation_safe_value): Add prototype.
7842         (default_speculation_safe_value): Add prototype.
7844 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
7846         * dump-context.h (dump_context::dump_loc): New decl.
7847         * dumpfile.c (dump_context::dump_loc): New member function.
7848         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
7849         and dump_gimple_stmt.
7850         (dump_context::dump_gimple_expr_loc): Likewise, using
7851         dump_gimple_expr.
7852         (dump_context::dump_generic_expr_loc): Likewise, using
7853         dump_generic_expr.
7854         (dump_context::dump_printf_loc_va): Likewise, using
7855         dump_printf_va.
7856         (dump_context::begin_scope): Explicitly using the global function
7857         "dump_loc", rather than the member function.
7859 2018-07-31  Martin Sebor  <msebor@redhat.com>
7861         PR tree-optimization/86741
7862         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
7864 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
7866         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
7867         depend on whether prefetch instructions will be emitted or not.
7868         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
7869         will be emitted or not.
7870         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
7871         (TARGET_SETMEM_PFD): New macros.
7873 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7875         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
7876         (NULL_STMT_VEC_INFO): Delete.
7877         (stmt_vec_info::operator*): Likewise.
7878         (stmt_vec_info::operator gimple *): Likewise.
7879         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
7880         of NULL_STMT_VEC_INFO.
7881         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7882         (vect_reassociating_reduction_p): Likewise.
7883         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
7884         (vectorizable_store): Likewise.
7885         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
7886         (vec_info::free_stmt_vec_infos): Likewise.
7888 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7890         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
7891         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
7893 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7895         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
7896         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
7897         (vec_info::free_stmt_vec_info): New private member functions.
7898         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
7899         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
7900         * tree-parloops.c (gather_scalar_reductions): Remove calls to
7901         set_stmt_vec_info_vec and free_stmt_vec_infos.
7902         * tree-vect-loop.c (_loop_vec_info): Remove call to
7903         set_stmt_vec_info_vec.
7904         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
7905         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
7906         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
7907         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
7908         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
7909         assignments in {vec_info::,}new_stmt_vec_info that are redundant
7910         with the clearing in the xcalloc.
7911         (stmt_vec_info_vec): Delete.
7912         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
7913         (vectorize_loops): Likewise.
7914         (vec_info::~vec_info): Remove argument from call to
7915         free_stmt_vec_infos.
7916         (vec_info::add_stmt): Remove vinfo argument from call to
7917         new_stmt_vec_info.
7919 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7921         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
7922         rather than a gimple stmt.
7923         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
7924         information for pattern statements when passed the original
7925         statement; instead wait to be passed the pattern statement itself.
7926         Don't call set_vinfo_for_stmt here.
7927         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
7928         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
7929         stmt_vec_infos here.
7930         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
7931         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
7932         stmt_vec_infos entry.
7934 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7936         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
7937         * tree-vectorizer.c (vec_info::replace_stmt): New function.
7938         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
7939         * tree-vect-stmts.c (vectorizable_call): Likewise.
7940         (vectorizable_simd_clone_call): Likewise.
7942 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7944         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
7945         * tree-vectorizer.c (vec_info::remove_stmt): New function.
7946         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
7947         * tree-vect-loop.c (vect_transform_loop): Likewise.
7948         * tree-vect-slp.c (vect_schedule_slp): Likewise.
7949         * tree-vect-stmts.c (vect_remove_stores): Likewise.
7951 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7953         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
7954         (vect_dr_stmt): Delete.
7955         * tree-vectorizer.c (vec_info::lookup_dr): New function.
7956         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
7957         of DR_VECT_AUX.
7958         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7959         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
7960         (vect_verify_datarefs_alignment, vect_peeling_supportable)
7961         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
7962         (vect_analyze_data_refs): Likewise.
7963         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
7964         argument.
7965         (vect_find_same_alignment_drs): Likewise.
7966         (vect_slp_analyze_node_dependences): Update calls accordingly.
7967         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
7968         instead of DR_VECT_AUX.
7969         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
7970         of a vector data references.  Use vec_info::lookup_dr instead of
7971         DR_VECT_AUX.
7972         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
7973         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
7974         instead of DR_VECT_AUX.
7976 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7978         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
7979         dr_vec_info.
7980         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
7981         accordingly.
7982         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7983         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
7984         (vect_gen_prolog_loop_niters): Likewise.
7986 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
7988         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
7989         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
7990         (vect_known_alignment_in_bytes, vect_dr_behavior)
7991         (vect_get_scalar_dr_size): Take references as dr_vec_infos
7992         instead of data_references.  Update calls to other routines for
7993         which the same change has been made.
7994         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
7995         dr_vec_infos instead of stmt_vec_infos.
7996         (vect_analyze_data_ref_dependence): Update call accordingly.
7997         (vect_slp_analyze_data_ref_dependence)
7998         (vect_record_base_alignments): Use DR_VECT_AUX.
7999         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
8000         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
8001         (vector_alignment_reachable_p, vect_get_data_access_cost)
8002         (vect_peeling_supportable, vect_analyze_group_access_1)
8003         (vect_analyze_group_access, vect_analyze_data_ref_access)
8004         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
8005         (vect_compile_time_alias, vect_small_gap_p)
8006         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
8007         (vect_supportable_dr_alignment): Take references as dr_vec_infos
8008         instead of data_references.  Update calls to other routines for
8009         which the same change has been made.
8010         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
8011         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
8012         (vect_slp_analyze_and_verify_node_alignment)
8013         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
8014         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8015         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
8016         above changes.
8017         (_vect_peel_info::dr): Replace with...
8018         (_vect_peel_info::dr_info): ...this new field.
8019         (vect_peeling_hash_get_most_frequent)
8020         (vect_peeling_hash_choose_best_peeling): Update accordingly.
8021         (vect_peeling_hash_get_lowest_cost):
8022         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
8023         routines for which the same change has been made.
8024         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
8025         data_reference.
8026         * tree-vect-loop-manip.c (get_misalign_in_elems)
8027         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
8028         above changes.
8029         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8030         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
8031         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
8032         (get_group_load_store_type, get_negative_load_store_type)
8033         (vect_get_data_ptr_increment, vectorizable_store)
8034         (vectorizable_load): Likewise.
8035         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
8036         Update calls to other routines for which the same change has been made.
8038 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8040         * tree-vectorizer.h (vec_info::move_dr): New member function.
8041         (dataref_aux): Rename to...
8042         (dr_vec_info): ...this and add "dr" and "stmt" fields.
8043         (_stmt_vec_info::dr_aux): Update accordingly.
8044         (_stmt_vec_info::data_ref_info): Delete.
8045         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
8046         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
8047         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
8048         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
8049         of data_ref.
8050         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
8051         (STMT_VINFO_DR_INFO): New macro.
8052         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
8053         (set_dr_misalignment): Update after rename of dataref_aux.
8054         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
8055         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
8056         initialization of STMT_VINFO_DATA_REF.
8057         * tree-vectorizer.c (vec_info::move_dr): New function.
8058         * tree-vect-patterns.c (vect_recog_bool_pattern)
8059         (vect_recog_mask_conversion_pattern)
8060         (vect_recog_gather_scatter_pattern): Use it.
8061         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
8062         the "dr" and "stmt" fields of dr_vec_info instead of
8063         STMT_VINFO_DATA_REF.
8065 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8067         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
8068         (is_pattern_stmt_p): Use it.
8069         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
8070         on pattern statements.
8072 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8074         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
8075         original stmt as a stmt_vec_info rather than a gimple stmt.
8076         (vect_pattern_recog_1): Take the statement directly as a
8077         stmt_vec_info, rather than via a gimple_stmt_iterator.
8078         Update call to vect_mark_pattern_stmts.
8079         (vect_pattern_recog): Update calls accordingly.
8081 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8083         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
8084         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
8085         a vect_def_type for the first argument.
8086         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
8087         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
8088         operand if it isn't defined by a vectorized statement.
8089         (vect_build_gather_load_calls): Remove the mask_dt argument and
8090         update calls to vect_get_vec_def_for_stmt_copy.
8091         (vectorizable_bswap): Likewise the dt argument.
8092         (vectorizable_call): Update calls to vectorizable_bswap and
8093         vect_get_vec_def_for_stmt_copy.
8094         (vectorizable_simd_clone_call, vectorizable_assignment)
8095         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
8096         (vectorizable_comparison): Update calls to
8097         vect_get_vec_def_for_stmt_copy.
8098         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
8099         vect_is_simple_use.
8100         (vect_get_loop_based_defs): Remove dt argument and update call
8101         to vect_get_vec_def_for_stmt_copy.
8102         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
8103         and vect_get_vec_def_for_stmt_copy.
8104         (vectorizable_load): Update calls to vect_build_gather_load_calls
8105         and vect_get_vec_def_for_stmt_copy.
8106         * tree-vect-loop.c (vect_create_epilog_for_reduction)
8107         (vectorizable_reduction, vectorizable_live_operation): Update calls
8108         to vect_get_vec_def_for_stmt_copy.
8110 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8112         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
8113         and gimple stmt arguments with a stmt_vec_info.
8114         (vect_record_base_alignments): Update calls accordingly.
8115         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
8116         and gimple stmt arguments with a stmt_vec_info.
8117         (vect_build_slp_tree_1): Remove vinfo argument and update call
8118         to vect_record_max_nunits.
8119         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
8120         and vect_record_max_nunits.
8122 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8124         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
8125         file and take a stmt_vec_info instead of a gimple stmt.
8126         (supportable_widening_operation, vect_finish_replace_stmt)
8127         (vect_finish_stmt_generation, vect_get_store_rhs)
8128         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
8129         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
8130         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
8131         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
8132         (vect_create_data_ref_ptr, bump_vector_ptr)
8133         (vect_permute_store_chain, vect_setup_realignment)
8134         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
8135         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
8136         (vectorizable_reduction, vectorizable_induction)
8137         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
8138         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
8139         than gimple stmts as arguments.
8140         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
8141         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
8142         (can_group_stmts_p, vect_check_gather_scatter)
8143         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
8144         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
8145         (vect_permute_load_chain, vect_shift_permute_load_chain)
8146         (vect_transform_grouped_load)
8147         (vect_record_grouped_load_vectors): Likewise.
8148         * tree-vect-loop.c (vect_fixup_reduc_chain)
8149         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
8150         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
8151         (vectorizable_reduction, vectorizable_induction)
8152         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
8153         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
8154         (vect_get_load_store_mask): Likewise.
8155         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
8156         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
8157         * tree-vect-stmts.c (vect_mark_relevant)
8158         (is_simple_and_all_uses_invariant)
8159         (exist_non_indexing_operands_for_use_p, process_use)
8160         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
8161         (vect_get_vec_def_for_operand, vect_get_vec_defs)
8162         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
8163         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
8164         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
8165         (get_negative_load_store_type, get_load_store_type)
8166         (vect_check_load_store_mask, vect_check_store_rhs)
8167         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
8168         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8169         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8170         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8171         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
8172         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
8173         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
8174         (supportable_widening_operation): Likewise.
8176 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8178         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
8179         a stmt_vec_info instead of a gcall.
8180         (vect_check_gather_scatter): Update call accordingly.
8181         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
8182         of a gphi.
8183         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
8184         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
8185         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
8186         instead of a gimple stmt.
8187         (vect_transform_loop): Update calls accordingly.
8188         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
8189         stmt_vec_infos instead of gimple stmts.
8190         (vect_analyze_slp_instance): Update use accordingly.
8191         * tree-vect-stmts.c (read_vector_array, write_vector_array)
8192         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
8193         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
8194         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
8195         (vect_gen_widened_results_half, vect_get_loop_based_defs)
8196         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
8197         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
8198         down to subroutines.
8200 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8202         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
8203         of the worklist from a vector of gimple stmts to a vector of
8204         stmt_vec_infos.
8205         * tree-vect-stmts.c (vect_mark_relevant, process_use)
8206         (vect_mark_stmts_to_be_vectorized): Likewise
8208 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8210         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
8211         statement before passing it to vect_analyze_stmt.
8212         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
8213         the chain of phi vector definitions.  Track the exit phi via its
8214         stmt_vec_info.
8215         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
8216         STMT_VINFO_REDUC_DEF.
8217         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
8218         stmt_vec_infos to handle the statement chains.
8219         (vect_get_slp_defs): Record the first statement in the node
8220         using a stmt_vec_info.
8221         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
8222         statements here and pass their stmt_vec_info down to subroutines.
8223         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
8224         down to vect_finish_stmt_generation.
8225         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
8226         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
8227         stmt_vec_infos to subroutines.
8228         (vect_remove_stores): Use stmt_vec_infos to handle the statement
8229         chains.
8231 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8233         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
8234         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
8235         (vect_permute_store_chain, vect_setup_realignment)
8236         (vect_permute_load_chain, vect_shift_permute_load_chain)
8237         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
8238         stmts internally, and when passing values to other vectorizer routines.
8239         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
8240         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
8241         (vect_analyze_loop_operations, get_initial_def_for_reduction)
8242         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
8243         (vectorizable_reduction, vectorizable_induction)
8244         (vectorizable_live_operation, vect_transform_loop_stmt)
8245         (vect_transform_loop): Likewise.
8246         * tree-vect-patterns.c (vect_reassociating_reduction_p)
8247         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
8248         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
8249         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8250         (vect_slp_analyze_node_operations_1): Likewise.
8251         * tree-vect-stmts.c (vect_mark_relevant, process_use)
8252         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
8253         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
8254         (vect_finish_stmt_generation_1, get_group_load_store_type)
8255         (get_load_store_type, vect_build_gather_load_calls)
8256         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8257         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8258         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8259         (vectorizable_store, vectorizable_load, vectorizable_condition)
8260         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
8261         (supportable_widening_operation): Likewise.
8262         (vect_get_vector_types_for_stmt): Likewise.
8263         * tree-vectorizer.h (vect_dr_behavior): Likewise.
8265 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8267         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8268         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
8269         (vect_permute_store_chain, vect_permute_load_chain)
8270         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
8271         repeated stmt_vec_info lookups.
8272         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
8273         (vect_update_ivs_after_vectorizer): Likewise.
8274         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
8275         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
8276         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
8277         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8278         (vect_bb_slp_scalar_cost): Likewise.
8279         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
8281 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8283         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
8284         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
8285         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
8286         to get gassigns and gcalls, rather than operating on generc gimple
8287         stmts.
8288         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
8289         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
8290         (vectorizable_load, vect_analyze_stmt): Likewise.
8291         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
8293 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8295         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
8296         return stmt_vec_infos rather than gimple stmts.  Do not accept
8297         null arguments.
8298         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
8299         of a gimple stmt.
8300         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
8301         Update use of get_later_stmt.
8302         (vect_get_constant_vectors): Update call accordingly.
8303         (vect_schedule_slp_instance): Likewise
8304         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
8305         (vect_slp_analyze_instance_dependence): Likewise.
8306         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
8308 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8310         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
8311         (stmt_info_for_cost::stmt_info): ...this new field.
8312         (add_stmt_costs): Update accordingly.
8313         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
8314         (vect_get_known_peeling_cost): Likewise.
8315         (vect_estimate_min_profitable_iters): Likewise.
8316         * tree-vect-stmts.c (record_stmt_cost): Likewise.
8318 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8320         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
8321         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
8322         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
8323         accordingly.
8324         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
8326 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8328         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
8329         a gimple stmt to a stmt_vec_info.
8330         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
8332 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8334         * tree-vectorizer.h (vec_info::grouped_stores): Change from
8335         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
8336         (_loop_vec_info::reduction_chains): Likewise.
8337         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
8338         accordingly.
8339         * tree-vect-slp.c (vect_analyze_slp): Likewise.
8341 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8343         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
8344         a gimple stmt to a stmt_vec_info.
8345         (_stmt_vec_info::next_element): Likewise.
8346         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
8347         (vect_slp_analyze_and_verify_node_alignment)
8348         (vect_analyze_group_access_1, vect_analyze_group_access)
8349         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
8350         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
8351         (vect_supportable_dr_alignment): Update accordingly.
8352         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
8353         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
8354         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
8355         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
8356         * tree-vect-slp.c (vect_build_slp_tree_1)
8357         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
8358         (vect_split_slp_store_group, vect_analyze_slp_instance)
8359         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
8360         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
8361         (get_group_load_store_type, get_load_store_type)
8362         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
8363         (vect_transform_stmt, vect_remove_stores): Likewise.
8365 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8367         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
8368         than a gimple stmt.
8369         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8370         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
8371         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
8372         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
8373         (vector_alignment_reachable_p, vect_get_data_access_cost)
8374         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
8375         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
8376         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
8377         (vect_analyze_group_access_1, vect_analyze_group_access)
8378         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
8379         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
8380         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
8381         result of vect_dr_stmt and use the stmt_vec_info instead of
8382         the associated gimple stmt.
8383         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8384         (vect_gen_prolog_loop_niters): Likewise.
8385         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8387 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8389         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
8390         to a vec<stmt_vec_info>.
8391         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
8392         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
8393         vec<stmt_vec_info>.
8394         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
8395         to a vec<stmt_vec_info>.
8396         (bst_traits::value_type, bst_traits::value_type): Likewise.
8397         (bst_traits::hash): Update accordingly.
8398         (vect_get_and_check_slp_defs): Change the stmts parameter from
8399         a vec<gimple *> to a vec<stmt_vec_info>.
8400         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
8401         (vect_build_slp_tree): Likewise.
8402         (vect_build_slp_tree_2): Likewise.  Update uses of
8403         SLP_TREE_SCALAR_STMTS.
8404         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
8405         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
8406         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
8407         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
8408         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
8409         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
8410         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
8411         (vect_get_constant_vectors, vect_get_slp_defs)
8412         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
8413         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
8414         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
8415         instead of gimple stmts.
8416         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
8417         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
8418         (vect_slp_analyze_instance_dependence): Update uses of
8419         SLP_TREE_SCALAR_STMTS.
8420         (vect_slp_analyze_and_verify_node_alignment): Likewise.
8421         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
8422         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
8423         (get_initial_defs_for_reduction): Likewise.
8424         (vect_create_epilog_for_reduction): Likewise.
8425         (vectorize_fold_left_reduction): Likewise.
8426         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
8427         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
8428         (can_vectorize_live_stmts): Likewise.
8430 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8432         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
8433         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
8434         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
8435         than gimple stmts.
8436         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
8437         of a gimple stmt.
8438         (gather_scalar_reductions): Update after above interface changes.
8439         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
8440         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
8441         than gimple stmts.
8442         (vect_force_simple_reduction): Likewise.
8443         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
8444         LOOP_VINFO_REDUCTIONS.
8445         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8447 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8449         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
8450         a gimple stmt to a stmt_vec_info.
8451         * tree-vect-loop.c (vect_active_double_reduction_p)
8452         (vect_force_simple_reduction, vectorizable_reduction): Update
8453         accordingly.
8455 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8457         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
8458         vec<gimple *> to a vec<stmt_vec_info>.
8459         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
8460         the reduction_phis argument from a vec<gimple *> to a
8461         vec<stmt_vec_info>.
8462         (vectorizable_reduction): Likewise the phis local variable that
8463         is passed to vect_create_epilog_for_reduction.  Update for new type
8464         of SLP_TREE_VEC_STMTS.
8465         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
8466         (vectorizable_live_operation): Likewise.
8467         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
8468         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
8470 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8472         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
8473         a gimple stmt to a stmt_vec_info.
8474         (vectorizable_condition, vectorizable_live_operation)
8475         (vectorizable_reduction, vectorizable_induction): Pass back the
8476         vectorized statement as a stmt_vec_info.
8477         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
8478         use of STMT_VINFO_VEC_STMT.
8479         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
8480         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
8481         as stmt_vec_infos rather than gimple stmts.
8482         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
8483         to a stmt_vec_info.
8484         (vectorizable_live_operation): Likewise.
8485         (vectorizable_reduction, vectorizable_induction): Likewise,
8486         updating use of STMT_VINFO_VEC_STMT.
8487         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
8488         of STMT_VINFO_VEC_STMT.
8489         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
8490         (vectorizable_simd_clone_call, vectorizable_conversion)
8491         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8492         (vectorizable_store, vectorizable_load, vectorizable_condition)
8493         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
8494         from a gimple stmt to a stmt_vec_info.
8495         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
8496         pointer to a stmt_vec_info to the vectorizable_* routines.
8498 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8500         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
8501         a gimple stmt to a stmt_vec_info.
8502         (is_pattern_stmt_p): Update accordingly.
8503         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
8504         (vect_record_grouped_load_vectors): Likewise.
8505         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
8506         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
8507         (vect_model_reduction_cost): Likewise.
8508         (vect_create_epilog_for_reduction): Likewise.
8509         (vectorizable_reduction, vectorizable_induction): Likewise.
8510         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8511         Return the stmt_vec_info for the pattern statement.
8512         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
8513         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
8514         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
8515         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
8516         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
8517         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
8518         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
8519         (free_stmt_vec_info, vect_is_simple_use): Likewise.
8521 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8523         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
8524         (vect_finish_stmt_generation): Likewise.
8525         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
8526         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
8527         (vect_build_gather_load_calls): Use the return value of the above
8528         functions instead of a separate call to vinfo_for_stmt.  Use narrow
8529         scopes for the input gimple stmt and wider scopes for the associated
8530         stmt_vec_info.  Use vec_info::lookup_def when setting these
8531         stmt_vec_infos from an SSA_NAME definition.
8532         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
8533         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
8534         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
8535         (vectorizable_store, vectorizable_load, vectorizable_condition)
8536         (vectorizable_comparison): Likewise.
8537         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
8538         (vectorizable_reduction): Likewise.
8540 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8542         * tree-vectorizer.h (vect_is_simple_use): Add an optional
8543         stmt_vec_info * parameter before the optional gimple **.
8544         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
8545         (process_use, vect_get_vec_def_for_operand_1): Update callers.
8546         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
8547         * tree-vect-loop.c (vectorizable_reduction): Likewise.
8548         (vectorizable_live_operation): Likewise.
8549         * tree-vect-patterns.c (type_conversion_p): Likewise.
8550         (vect_look_through_possible_promotion): Likewise.
8551         (vect_recog_rotate_pattern): Likewise.
8552         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
8554 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8556         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
8557         a typedef to a wrapper class.
8558         (NULL_STMT_VEC_INFO): New macro.
8559         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
8560         (stmt_vec_info::operator*): New function.
8561         (stmt_vec_info::operator gimple *): Likewise.
8562         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
8563         (add_stmt_costs): Likewise.
8564         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
8565         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
8566         (vect_get_known_peeling_cost): Likewise.
8567         (vect_estimate_min_profitable_iters): Likewise.
8568         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8569         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
8570         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
8571         (vectorizable_store, free_stmt_vec_infos): Likewise.
8572         (new_stmt_vec_info): Change return type of xcalloc to
8573         _stmt_vec_info *.
8575 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8577         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
8578         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
8579         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
8580         a single_imm_use-based sequence.
8581         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
8583 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8585         * tree-vectorizer.h (vec_info::lookup_def): Declare.
8586         * tree-vectorizer.c (vec_info::lookup_def): New function.
8587         * tree-vect-patterns.c (vect_get_internal_def): Use it.
8588         (vect_widened_op_tree): Likewise.
8589         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
8590         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
8591         (vectorizable_reduction): Likewise.
8592         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
8593         of a gimple *.
8594         (vect_is_slp_reduction): Update calls accordingly.  Use
8595         vec_info::lookup_def.
8596         (vect_is_simple_reduction): Likewise
8597         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
8599 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8601         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
8602         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
8603         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
8604         of vinfo_for_stmt.
8605         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
8606         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
8607         (vect_update_vf_for_slp, vect_analyze_loop_operations)
8608         (vect_is_slp_reduction, vectorizable_induction)
8609         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
8610         * tree-vect-patterns.c (vect_init_pattern_stmt):
8611         (vect_determine_min_output_precision_1, vect_determine_precisions)
8612         (vect_pattern_recog): Likewise.
8613         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
8614         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
8615         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
8616         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
8617         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
8618         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
8619         info field from a loop to a loop_vec_info.
8621 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8623         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
8624         (vec_info::add_stmt): Declare.
8625         * tree-vectorizer.c (vec_info::add_stmt): New function.
8626         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
8627         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
8628         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
8629         (vectorizable_induction): Likewise.
8630         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
8631         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
8632         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
8633         (vectorizable_load): Likewise.
8634         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
8635         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
8636         (vect_recog_gather_scatter_pattern): Likewise.
8637         (append_pattern_def_seq): Likewise.  Remove a check that is
8638         performed by add_stmt itself.
8640 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8642         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
8643         which make_ssa_name was called with new_stmt before new_stmt
8644         had been created.
8646 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8648         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
8649         split out from...
8650         (vect_is_slp_reduction): ...here...
8651         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
8652         that are already known to be false.
8654 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8656         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
8657         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
8658         STMT_VINFO_NUM_SLP_USES when it's true.
8659         (vect_free_slp_instance): Add a final_p parameter and pass it to
8660         vect_free_slp_tree.
8661         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
8662         (vect_analyze_slp_instance): Likewise.
8663         (vect_slp_analyze_operations): Likewise.
8664         (vect_slp_analyze_bb_1): Likewise.
8665         * tree-vectorizer.c (vec_info): Likewise.
8666         * tree-vect-loop.c (vect_transform_loop): Likewise.
8668 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8670         * tree-vect-loop.c (vectorizable_reduction): Assert that the
8671         function is not called for second and subsequent members of
8672         a reduction group.
8674 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
8676         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
8677         cases for nested loops from here to ...
8678         (vect_create_epilog_for_reduction): ...here.  Only call
8679         vect_is_simple_use for inner-loop reductions.
8681 2018-07-31  Martin Liska  <mliska@suse.cz>
8683         PR gcov-profile/85338
8684         PR gcov-profile/85350
8685         PR gcov-profile/85372
8686         * profile.c (struct location_triplet): New.
8687         (struct location_triplet_hash): Likewise.
8688         (output_location): Do not output a BB that
8689         is already recorded for a line.
8690         (branch_prob): Use streamed_locations.
8692 2018-07-31  Martin Liska  <mliska@suse.cz>
8694         PR gcov-profile/85370
8695         * coverage.c (coverage_begin_function): Do not mark target
8696         clones as artificial functions.
8698 2018-07-31  Martin Liska  <mliska@suse.cz>
8700         PR gcov-profile/83813
8701         PR gcov-profile/84758
8702         PR gcov-profile/85217
8703         PR gcov-profile/85332
8704         * profile.c (branch_prob): Do not record GOTO expressions
8705         for GIMPLE statements which locations are already streamed.
8707 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8709         * gcc.c (handle_spec_function): Accept a soft_matched_part
8710         argument, as do_spec_1.  Pass it down to ...
8711         (eval_spec_function): Accept a soft_matched_part argument,
8712         and pass it down to ...
8713         (do_spec_2): Accept a soft_matched_part argument, and pass
8714         it down to do_spec_1.
8715         (do_spec_1): Pass soft_matched_part to handle_spec_function.
8716         (handle_braces): Update call to handle_spec_function.
8717         (driver::set_up_specs): Update calls to do_spec_2.
8718         (compare_debug_dump_opt_spec_function): Likewise.
8719         (compare_debug_self_opt_spec_function): Likewise.
8721 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8723         * common.opt (nolibc): New option.
8724         * doc/invoke.texi (Link Options): Document it.
8725         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
8726         * config/alpha/linux.h: Likewise.
8727         * config/arc/elf.h: Likewise.
8728         * config/arm/uclinux-elf.h: Likewise.
8729         * config/arm/unknown-elf.h: Likewise.
8730         * config/avr/avrlibc.h: Likewise.
8731         * config/bfin/bfin.h: Likewise.
8732         * config/bfin/linux.h: Likewise.
8733         * config/bfin/uclinux.h: Likewise.
8734         * config/darwin.h: Likewise.
8735         * config/darwin10.h: Likewise.
8736         * config/darwin12.h: Likewise.
8737         * config/gnu-user.h: Likewise.
8738         * config/lm32/uclinux-elf.h: Likewise.
8739         * config/pa/pa-hpux11.h: Likewise.
8740         * config/pa/pa64-hpux.h: Likewise.
8741         * config/sparc/sparc.h: Likewise.
8743 2018-07-31  Olivier Hainque  <hainque@adacore.com>
8745         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
8746         undefined variables.
8748 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
8750         PR target/86640
8751         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
8752         instead of GEN_INT.
8754 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8756         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
8757         terminated string literal.
8759 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
8761         PR rtl-optimization/85160
8762         * combine.c (is_just_move): New function.
8763         (try_combine): Allow combining two instructions into two if neither of
8764         the original instructions was a move.
8766 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
8768         PR target/86673
8769         * doc/extend.texi (Global Register Variables): Discourage use of type
8770         qualifiers.
8771         (Local Register Variables): Likewise.
8773 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
8775         PR tree-optimization/86506
8776         * hwint.h (ceil_log2): Resync with hwint.c implementation.
8778 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
8780         PR target/86547
8781         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
8782         hard_regno, make sure no insn between `from` and `to` clobbers it.
8784 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
8785             Tom de Vries  <tdevries@suse.de>
8787         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
8788         (PTX_DEFAULT_RUNTIME_DIM): ... this.
8789         (nvptx_goacc_validate_dims): Set default worker and gang dims to
8790         PTX_DEFAULT_RUNTIME_DIM.
8791         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
8793 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
8795         * config/pa/pa.c (pa_output_addr_vec): Align address table.
8796         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
8797         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
8799 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
8801         * config/rs6000/constraints.md (wG constraint): Delete, no longer
8802         used.
8803         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
8804         predicate to reflect toc fusion has been deleted.
8805         (toc_fusion_mem_raw): Delete, no longer used.
8806         (toc_fusion_mem_wrapped): Likewise.
8807         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
8808         fusion mask bit.
8809         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
8810         Delete, no longer used.
8811         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
8812         meant to be used for toc fusion.
8813         (rs6000_debug_print_mode): Delete toc fusion debugging.
8814         (rs6000_debug_reg_global): Likewise.
8815         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
8816         fusion and secondary reload support that were never used.
8817         (rs6000_option_override_internal): Delete TOC fusion, that was only
8818         partially defined, and it did not work unless you also used the
8819         -mcmodel= switch.
8820         (rs6000_legitimate_address_p): Delete TOC fusion support.
8821         (rs6000_opt_masks): Likewise.
8822         (fusion_wrap_memory_address): Delete function, no longer used.
8823         (fusion_split_address); Delete TOC fusion support.
8824         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
8825         longer used with toc fusion being deleted.
8826         (TARGET_TOC_FUSION_FP): Likewise.
8827         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
8828         UNSPEC.
8829         (toc fusion spliter): Delete TOC fusion support.
8830         (toc_fusionload_<mode>): Likewise.
8831         (toc_fusionload_di): Likewise.
8832         (fusion_gpr_load_<mode>): Delete generator function, this insn no
8833         longer needs to be named.  Rename predicate to delete TOC fusion.
8834         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
8835         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
8836         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
8837         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
8838         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
8840 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8842         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
8843         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
8844         __int128 in built-in function prototypes.
8845         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
8846         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
8848 2018-07-27  Martin Sebor  <msebor@redhat.com>
8850         PR tree-optimization/86696
8851         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
8852         types, including enums.
8853         (handle_char_store): Be prepared for the above function to fail.
8855 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
8857         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
8858         when optimization level is lower than 2 or optimize for size.
8860 2018-07-26  Martin Sebor  <msebor@redhat.com>
8862         PR tree-optimization/86043
8863         PR tree-optimization/86042
8864         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
8865         (get_string_cst_length): Rename...
8866         (get_min_string_length): ...to this.  Add argument.
8867         (handle_char_store): Extend to handle multi-character stores by
8868         MEM_REF.
8869         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
8870         * tree.h (initializer_zerop): Add argument.
8872 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
8874         PR middle-end/86660
8875         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
8876         declare target to variables if they have always,{to,from,tofrom} map
8877         kinds.
8879 2018-07-26  Martin Liska  <mliska@suse.cz>
8881         PR lto/86548
8882         * lto-wrapper.c: Add linker_output as prefix
8883         for ltrans_output_file.
8885 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
8887         PR rtl-optimization/85805
8888         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
8889         value for hard registers if that was written in the same mode.
8891 2018-07-26  Martin Liska  <mliska@suse.cz>
8893         PR gcov-profile/86536
8894         * gcov.c (format_gcov): Use printf format %.*f directly
8895         and do not handle special values.
8897 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8899         * common/config/arc/arc-common.c (arc_option_optimization_table):
8900         Update default optimizations for size.
8902 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8904         * config/arc/arc.md (movsf_insn): Add short instruction selection.
8905         * config/arc/constraints.md (CfZ): New constraint.
8906         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
8907         (subsf3_fpu): Likewise.
8908         (cmpsf_fpu): Likewise.
8909         (cmpsf_fpu_uneq): Likewise.
8911 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8913         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
8914         accesses as well.
8915         (arc_is_uncached_mem_p): uncached applies to both the variable and
8916         the pointer.
8918 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
8920         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
8921         register names.
8923 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
8925         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
8926         field "m_scopes" from vec to auto_vec.
8928 2018-07-25  Martin Liska  <mliska@suse.cz>
8930         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
8931         return type.
8933 2018-07-25  Richard Biener  <rguenther@suse.de>
8935         PR debug/86654
8936         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
8937         special wrt context_die late.
8938         (gen_subprogram_die): Re-use DIEs in local scope.
8940 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
8942         PR tree-optimization/86644
8943         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
8945 2018-07-25  Martin Liska  <mliska@suse.cz>
8947         PR middle-end/86645
8948         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
8949         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
8951 2018-07-25  Martin Liska  <mliska@suse.cz>
8953         PR sanitizer/79635
8954         * params.def: Explain ASan abbreviation and provide
8955         a documentation link.
8957 2018-07-24  Martin Sebor  <msebor@redhat.com>
8959         PR tree-optimization/86622
8960         PR tree-optimization/86532
8961         * builtins.h (string_length): Declare.
8962         * builtins.c (c_strlen): Correct handling of non-constant offsets.
8963         (check_access): Be prepared for non-constant length ranges.
8964         (string_length): Make extern.
8965         * expr.c (string_constant): Only handle the minor non-constant
8966         array index.  Use string_constant to compute the length of
8967         a generic string constant.
8969 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
8971         PR tree-optimization/86618
8972         * tree-vect-stmts.c (vectorizable_call): Don't take the address
8973         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
8975 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
8977         PR tree-optimization/86636
8978         * json.cc (json::object::set): Fix comment.  Add assertions.
8979         (json::array::append): Move here from json.h.  Add comment and an
8980         assertion.
8981         (json::string::string): Likewise.
8982         * json.h (json::array::append): Move to json.cc.
8983         (json::string::string): Likewise.
8984         * optinfo-emit-json.cc
8985         (optrecord_json_writer::impl_location_to_json): Assert that we
8986         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
8987         wrapper around it.  Expand the location once, rather than three
8988         times.
8989         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
8990         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
8991         wrappers.
8992         (optrecord_json_writer::optinfo_to_json): Likewise, in four
8993         places.  Fix some overlong lines.
8995 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
8997         * config/aarch64/aarch64-simd.md
8998         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
8999         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
9000         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
9001         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
9002         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
9003         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
9004         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
9005         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
9006         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
9008 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
9010         PR middle-end/86627
9011         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
9012         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
9013         and abs_d == d, do the power of two handling if profitable.
9015 2018-07-24  Richard Biener  <rguenther@suse.de>
9017         * match.pd: Add BIT_FIELD_REF canonicalizations.
9019 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9021         PR c/86617
9022         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
9024 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9026         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
9027         terminated STRING_CST object.
9029 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9031         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
9033 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
9035         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
9036         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
9037         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
9038         the elements into a register.
9039         (rs6000_split_v4si_init_di_reg): Delete.
9040         (rs6000_split_v4si_init): Delete.
9041         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
9042         (vsx_init_v4si): Rewrite as a define_expand.
9044 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
9046         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
9047         zero_extend argument from memory): New.
9049 2018-07-22  Martin Sebor  <msebor@redhat.com>
9051         PR bootstrap/86621
9052         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
9053         diagnosing calls with unknown arguments unless -Walloca-larger-than
9054         is restricted to less than PTRDIFF_MAX bytes.
9056 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
9058         * doc/gcov.texi (Invoking Gcov): Editorial changes.
9060 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
9062         * pretty-print.c (text_info::set_location): Remove redundant
9063         "line_table" parameter from call to rich_location::set_range.
9065 2018-07-20  Martin Sebor  <msebor@redhat.com>
9067         PR middle-end/82063
9068         * builtins.c (expand_builtin_alloca): Adjust.
9069         * calls.c (alloc_max_size): Simplify.
9070         * cgraphunit.c (cgraph_node::expand): Adjust.
9071         * common.opt (larger_than_size, warn_frame_larger_than): Remove
9072         variables.
9073         (frame_larger_than_size): Same.
9074         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
9075         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
9076         Initialize.
9077         * doc/invoke.texi (GCC Command Options): Document option arguments.
9078         Explain byte-size arguments and suffixes.
9079         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
9080         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
9081         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
9082         * doc/options.texi (UInteger): Expand.
9083         (Host_Wide_Int, ByteSize): Document new properties.
9084         * final.c (final_start_function_1): Include sizes in an error message.
9085         * function.c (frame_offset_overflow): Same.
9086         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
9087         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
9088         Diagnose unbounded alloca calls only for limits of less than
9089         PTRDIFF_MAX.
9090         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
9091         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
9092         for alloca(0).
9093         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
9094         only for limits of less than PTRDIFF_MAX.
9095         * langhooks-def.h (lhd_handle_option): Change function argument
9096         to HOST_WIDE_INT.
9097         * langhooks.c (lhd_handle_option): Same.
9098         * langhooks.h (handle_option): Same.
9099         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
9100         ByteSize flags.
9101         (var_type, var_type_struct): Same.
9102         (var_set): Handle ByteSize flag.
9103         * optc-gen.awk: Add comments to output to ease debugging.  Make
9104         use of HOST_WIDE_INT where appropriate.
9105         * opts-gen-save.awk:  Use %lx to format unsigned long.
9106         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
9107         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
9108         arguments.  Parse bytes-size suffixes.
9109         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
9110         (enum_value_to_arg): Same.
9111         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
9112         (handle_option): Adjust.
9113         (generate_option): Change function argument to HOST_WIDE_INT.
9114         (cmdline_handle_error): Adjust.
9115         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
9116         (set_option): Change function argument to HOST_WIDE_INT.
9117         (option_enabled): Handle cl_host_wide_int.
9118         (get_option_state): Handle CLVC_SIZE.
9119         (control_warning_option): Same.
9120         * opts.c (common_handle_option): Change function argument to
9121         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
9122         OPT_Wvla_larger_than_.
9123         * opts.h (enum cl_var_type): Add an enumerator.
9124         * stor-layout.c (layout_decl): Print a more meaningful warning.
9125         * toplev.c (output_stack_usage): Adjust.
9127 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
9129         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
9130         call to inline_expand_builtin_string_cmp.
9131         (expand_builtin_strcmp): Likewise.
9132         (expand_builtin_strncmp): Likewise.
9133         (inline_string_cmp): Delete the last parameter, change char_type_node
9134         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
9135         two operands.
9136         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
9137         the inlining expansion on target where the type of the call has same or
9138         narrower precision than unsigned char.
9140 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
9142         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
9143         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
9144         * common.opt (fsave-optimization-record): New option.
9145         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
9146         * doc/invoke.texi (-fsave-optimization-record): New option.
9147         * dumpfile.c: Include "optinfo-emit-json.h".
9148         (struct kv_pair): Move to coretypes.h.
9149         (optgroup_options): Make non-static.
9150         (dump_context::end_scope): Call
9151         optimization_records_maybe_pop_dump_scope.
9152         * dumpfile.h (optgroup_options): New decl.
9153         * json.cc: New file.
9154         * json.h: New file.
9155         * optinfo-emit-json.cc: New file.
9156         * optinfo-emit-json.h: New file.
9157         * optinfo.cc: Include "optinfo-emit-json.h".
9158         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
9159         (optinfo_enabled_p): Check optimization_records_enabled_p.
9160         (optinfo_wants_inlining_info_p): Likewise.
9161         * optinfo.h: Update comment.
9162         * profile-count.c (profile_quality_as_string): New function.
9163         * profile-count.h (profile_quality_as_string): New decl.
9164         (profile_count::quality): New accessor.
9165         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
9166         and optinfo_emit_json_cc_tests.
9167         * selftest.h (selftest::json_cc_tests): New decl.
9168         (selftest::optinfo_emit_json_cc_tests): New decl.
9169         * toplev.c: Include "optinfo-emit-json.h".
9170         (compile_file): Call optimization_records_finish.
9171         (do_compile): Call optimization_records_start.
9172         * tree-ssa-live.c: Include optinfo.h.
9173         (remove_unused_scope_block_p): Retain inlining information if
9174         optinfo_wants_inlining_info_p returns true.
9176 2018-07-20  Richard Biener  <rguenther@suse.de>
9178         PR debug/86585
9179         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
9180         to cover -flto-partition=none.
9182 2018-07-20  Martin Liska  <mliska@suse.cz>
9184         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
9185         (get_decl_source_range): Remove unused function.
9187 2018-07-20  Richard Biener  <rguenther@suse.de>
9189         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
9190         (struct vn_phi_s): Likewise.
9191         (struct vn_reference_s): Likewise.
9192         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
9193         for searching the slot of an entry known to be in the hash itself.
9194         (vn_phi_hasher::equal): Likewise.
9195         (vn_reference_hasher::equal): Likewise.
9196         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
9197         globals.
9198         (optimistic_info, current_info): Remove, keeping only valid_info.
9199         (vn_reference_lookup_1): Remove fallback lookup.
9200         (vn_reference_lookup_2): Likewise.
9201         (vn_nary_op_lookup_1): Likewise.
9202         (vn_phi_lookup): Likewise.
9203         (vn_nary_build_or_lookup_1): Make sure to not chain the built
9204         hash element.
9205         (vn_reference_insert): Adjust, chain the inserted hash element
9206         at last_inserted_ref.
9207         (vn_reference_insert_pieces): Likewise.
9208         (visit_reference_op_call): Likewise.
9209         (vn_nary_op_insert_into): Chain the inserted hash element at
9210         last_inserted_nary.
9211         (vn_nary_op_insert_pieces): Adjust.
9212         (vn_nary_op_insert): Likewise.
9213         (vn_nary_op_insert_stmt): Likewise.
9214         (vn_phi_insert): Adjust, chain the inserted hash element at
9215         last_inserted_phi.
9216         (process_scc): Remove clearing and copying the optimistic
9217         table.  Instead remove elements inserted during an optimistic
9218         iteration from the single table we maintain.
9219         (init_scc_vn): Adjust.
9220         (free_scc_vn): Likewise.
9221         (sccvn_dom_walker::record_cond): Likewise.
9222         (sccvn_dom_walker::after_dom_children): Likewise.
9224 2018-07-19  Martin Sebor  <msebor@redhat.com>
9226         PR tree-optimization/84047
9227         PR tree-optimization/83776
9228         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
9229         (check_array_bounds): Call it.
9231 2018-07-19  Martin Sebor  <msebor@redhat.com>
9233         * align.h (align_flags): Use member initialization.
9235 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
9237         * Makefile.in (OBJS): Add optinfo.o.
9238         * coretypes.h (class symtab_node): New forward decl.
9239         (struct cgraph_node): New forward decl.
9240         (class varpool_node): New forward decl.
9241         * dump-context.h: New file.
9242         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
9243         "tree-pass.h".
9244         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
9245         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
9246         (set_alt_dump_file): Likewise.
9247         (dump_context::~dump_context): New dtor.
9248         (dump_gimple_stmt): Move implementation to...
9249         (dump_context::dump_gimple_stmt): ...this new member function.
9250         Add the stmt to any pending optinfo, creating one if need be.
9251         (dump_gimple_stmt_loc): Move implementation to...
9252         (dump_context::dump_gimple_stmt_loc): ...this new member function.
9253         Start a new optinfo and add the stmt to it.
9254         (dump_gimple_expr): Move implementation to...
9255         (dump_context::dump_gimple_expr): ...this new member function.
9256         Add the stmt to any pending optinfo, creating one if need be.
9257         (dump_gimple_expr_loc): Move implementation to...
9258         (dump_context::dump_gimple_expr_loc): ...this new member function.
9259         Start a new optinfo and add the stmt to it.
9260         (dump_generic_expr): Move implementation to...
9261         (dump_context::dump_generic_expr): ...this new member function.
9262         Add the tree to any pending optinfo, creating one if need be.
9263         (dump_generic_expr_loc): Move implementation to...
9264         (dump_context::dump_generic_expr_loc): ...this new member
9265         function.  Add the tree to any pending optinfo, creating one if
9266         need be.
9267         (dump_printf): Move implementation to...
9268         (dump_context::dump_printf_va): ...this new member function.  Add
9269         the text to any pending optinfo, creating one if need be.
9270         (dump_printf_loc): Move implementation to...
9271         (dump_context::dump_printf_loc_va): ...this new member function.
9272         Start a new optinfo and add the stmt to it.
9273         (dump_dec): Move implementation to...
9274         (dump_context::dump_dec): ...this new member function.  Add the
9275         value to any pending optinfo, creating one if need be.
9276         (dump_context::dump_symtab_node): New member function.
9277         (dump_context::get_scope_depth): New member function.
9278         (dump_context::begin_scope): New member function.
9279         (dump_context::end_scope): New member function.
9280         (dump_context::ensure_pending_optinfo): New member function.
9281         (dump_context::begin_next_optinfo): New member function.
9282         (dump_context::end_any_optinfo): New member function.
9283         (dump_context::s_current): New global.
9284         (dump_context::s_default): New global.
9285         (dump_scope_depth): Delete global.
9286         (dumpfile_ensure_any_optinfo_are_flushed): New function.
9287         (dump_symtab_node): New function.
9288         (get_dump_scope_depth): Reimplement in terms of dump_context.
9289         (dump_begin_scope): Likewise.
9290         (dump_end_scope): Likewise.
9291         (selftest::temp_dump_context::temp_dump_context): New ctor.
9292         (selftest::temp_dump_context::~temp_dump_context): New dtor.
9293         (selftest::verify_item): New function.
9294         (ASSERT_IS_TEXT): New macro.
9295         (ASSERT_IS_TREE): New macro.
9296         (ASSERT_IS_GIMPLE): New macro.
9297         (selftest::test_capture_of_dump_calls): New test.
9298         (selftest::dumpfile_c_tests): Call it.
9299         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
9300         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
9301         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
9302         descriptive comment.
9303         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
9304         (dump_node, dump_bb): Move these unrelated decls.
9305         (class dump_manager): Add leading comment.
9306         * optinfo.cc: New file.
9307         * optinfo.h: New file.
9309 2018-07-19  Michael Collison  <michael.collison@arm.com>
9310             Richard Henderson <rth@redhat.com>
9312         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
9313         (subti): Handle op1 zero.
9314         (subvti4, usub4ti4): New.
9315         (*sub<GPI>3_compare1_imm): New.
9316         (sub<GPI>3_carryinCV): New.
9317         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
9318         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
9320 2018-07-19  Michael Collison  <michael.collison@arm.com>
9321             Richard Henderson <rth@redhat.com>
9323         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
9324         (addti3): Create simpler code if low part is already known to be 0.
9325         (addvti4, uaddvti4): New.
9326         (*add<GPI>3_compareC_cconly_imm): New.
9327         (*add<GPI>3_compareC_cconly): New.
9328         (*add<GPI>3_compareC_imm): New.
9329         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
9330         handle constants within this pattern..
9331         (*add<GPI>3_compareV_cconly_imm): New.
9332         (*add<GPI>3_compareV_cconly): New.
9333         (*add<GPI>3_compareV_imm): New.
9334         (add<GPI>3_compareV): New.
9335         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
9336         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
9337         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
9338         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
9339         with 'comparison' operator.
9340         (*add<GPI>3_compareV_cconly_imm): Ditto.
9341         (*add<GPI>3_compareV_cconly): Ditto.
9342         (*add<GPI>3_compareV_imm): Ditto.
9343         (add<GPI>3_compareV): Ditto.
9344         (add<mode>3_carryinC): Ditto.
9345         (*add<mode>3_carryinC_zero): Ditto.
9346         (*add<mode>3_carryinC): Ditto.
9347         (add<mode>3_carryinV): Ditto.
9348         (*add<mode>3_carryinV_zero): Ditto.
9349         (*add<mode>3_carryinV): Ditto.
9351 2018-07-19  Michael Collison  <michael.collison@arm.com>
9352             Richard Henderson <rth@redhat.com>
9354         * config/aarch64/aarch64-modes.def (CC_V): New.
9355         * config/aarch64/aarch64-protos.h
9356         (aarch64_addti_scratch_regs): Declare
9357         (aarch64_subvti_scratch_regs): Declare.
9358         (aarch64_expand_subvti): Declare.
9359         (aarch64_gen_unlikely_cbranch): Declare
9360         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
9361         for signed overflow using CC_Vmode.
9362         (aarch64_get_condition_code_1): Handle CC_Vmode.
9363         (aarch64_gen_unlikely_cbranch): New function.
9364         (aarch64_addti_scratch_regs): New function.
9365         (aarch64_subvti_scratch_regs): New function.
9366         (aarch64_expand_subvti): New function.
9368 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
9370         * config/aarch64/aarch64-option-extensions.def: New entry for profile
9371         extension.
9372         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
9373         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
9374         extension.
9376 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
9378         PR target/83009
9379         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9380         address check not strict.
9382 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
9384         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
9385         Umq with Umn.
9386         (store_pair_lanes<mode>): Likewise.
9387         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
9388         enum value 'ADDR_QUERY_LDP_STP_N'.
9389         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
9390         (aarch64_print_address_internal): Add declaration.
9391         (aarch64_print_ldpstp_address): Remove.
9392         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
9393         (aarch64_print_operand): Change printing of 'y'.
9394         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
9395         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
9396         'true' rather than '1'.
9397         * config/aarch64/constraints.md (Uml): Likewise.
9398         (Uml): Rename to Umn.
9399         (Umq): Remove.
9401 2018-07-19  Richard Biener  <rguenther@suse.de>
9403         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
9404         a trailing array.
9405         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
9406         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
9407         (vn_reference_hasher): Likewise.
9408         (struct vn_tables_s): Remove obstack and alloc-pool members.
9409         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
9410         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
9411         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
9412         (vn_reference_insert_pieces): Likewise.
9413         (alloc_vn_nary_op_noinit): Adjust.
9414         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
9415         (vn_phi_eq): Adjust.
9416         (shared_lookup_phiargs): Remove.
9417         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
9418         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
9419         (visit_reference_op_call): Likewise.
9420         (copy_nary, copy_phi, copy_reference): Remove.
9421         (process_scc): Rewind the obstack when iterating.  Do not
9422         copy the elements to valid_info but just move them from one
9423         hashtable to the other.
9424         (allocate_vn_table): Adjust.
9425         (free_vn_table): Likewise.
9426         (init_scc_vn): Likewise.
9427         (free_scc_vn): Likewise.
9429 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
9431         PR target/86560
9432         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
9433         indirect_return as function type attribute.
9434         (ix86_attribute_table): Change indirect_return to function
9435         type attribute.
9436         * doc/extend.texi: Update indirect_return attribute.
9438 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
9440         * wide-int.h (widest2_int): New.
9441         * gimple-fold.c (arith_overflowed_p): Use it.
9442         * tree.h (widest2_int_cst): New.
9443         * tree-vrp.c (wide_int_binop_overflow): Rename from
9444         vrp_int_const_binop.
9445         Rewrite to work on trees.
9446         (extract_range_from_multiplicative_op_1): Abstract code to...
9447         (wide_int_range_min_max): ...here.
9448         (wide_int_range_cross_product): ...and here.
9449         (extract_range_from_binary_expr_1): Abstract overflow code to...
9450         (wide_int_range_mult_wrapping): ...here.
9451         * tree-vrp.h (wide_int_range_cross_product): New.
9452         (wide_int_range_mult_wrapping): New.
9454 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
9455             Julia Koval  <julia.koval@intel.com>
9457         * config/i386/x86-tune-costs.h (skylake_memcpy,
9458         skylake_memset): Replace rep_prefix with unrolling for size 512.
9460 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
9462         PR middle-end/86544
9463         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
9464         comparision with EQ_EXPR in last stmt.
9466 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9468         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
9469         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
9470         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
9471         previously known as "PowerPC AltiVec Built-in Functions".  Move
9472         some material to new subsubsections "PowerPC AltiVec Built-in
9473         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
9474         ISA 2.07".
9475         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
9476         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
9477         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
9478         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
9480 2018-07-18  Richard Biener  <rguenther@suse.de>
9482         PR tree-optimization/86557
9483         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
9484         EXACT_DIV_EXPR.
9486 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
9488         * config/s390/s390.c (s390_function_profiler): Generate CFI.
9490 2018-07-17  Jeff Law  <law@redhat.com>
9492         * config/arm/arm.c (get_label_padding): Update for recent
9493         changes to label_to_alignment.
9495         PR tree-optimization/86010
9496         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
9498         * config/mips/mips.c (vr4130_align_insns): Update for recent
9499         changes to label_to_alignment.
9501         * config/frv/frv.c (frv_label_align): Update for recent changes
9502         to label_to_alignment.
9504         * config/nios2/nios2.c (nios2_label_align): Update for recent
9505         changes which dropped ALIGN_LABELS_LOG.
9507 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
9509         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
9510         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
9512 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
9514         * config/arc/arc.c (arc_label_align): Use align_labels instead of
9515         deprecated align_labels_log.
9517 2018-07-17  Richard Biener  <rguenther@suse.de>
9519         PR lto/86456
9520         * dwarf2out.c (init_sections_and_labels): Always generate
9521         a debug_line_str_section for early LTO debug.
9522         (dwarf2out_finish): Reset debug_line_str_hash output early.
9523         Bump counter for extra dwarf5 .debug_loc labels to not conflict
9524         with fat LTO part.
9525         (dwarf2out_early_finish): Output debug_line_str.
9527 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
9529         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
9530         index register on z196 or later.
9532 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
9534         * config/s390/s390.c (s390_default_align): Set default function
9535         alignment to 16.
9536         (s390_override_options_after_change): Call s390_default align.
9537         (s390_option_override_internal): Call s390_default_align.
9538         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
9540 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
9542         PR middle-end/86542
9543         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
9544         to _looptemp_ clauses, other than the first two.
9546 2018-07-17  Martin Liska  <mliska@suse.cz>
9548         * opts.c: Do not enable OPT_falign_* for -Os.
9550 2018-07-17  Martin Liska  <mliska@suse.cz>
9552         * align.h (MAX_CODE_ALIGN): New.
9553         (MAX_CODE_ALIGN_VALUE): New.
9554         * common/config/i386/i386-common.c (ix86_handle_option):
9555         (MAX_CODE_ALIGN): Moved to align.h.
9556         * final.c (MAX_CODE_ALIGN): Likewise.
9557         * opts.c (parse_and_check_align_values):
9558         (MAX_CODE_ALIGN): Likewise.
9559         (MAX_CODE_ALIGN_VALUE): Likewise.
9561 2018-07-17  Martin Liska  <mliska@suse.cz>
9563         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
9564         in order to fulfil coding style.
9565         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
9566         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9567         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9568         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
9569         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
9570         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
9571         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
9572         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
9573         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9575 2018-07-17  Martin Liska  <mliska@suse.cz>
9577         * align.h: New file.
9578         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
9579         directly.
9580         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
9581         align_flags of label_to_alignment.
9582         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
9583         align_flags class.
9584         * config/m68k/m68k.c: Do not use removed align_labels_value and
9585         align_loops_value.
9586         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
9587         (LOOP_ALIGN): Likewise.
9588         (LABEL_ALIGN): Likewise.
9589         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
9590         Remove not used macro.
9591         (rs6000_loop_align): Change return type to align_flags.
9592         (rs6000_loop_align_max_skip): Remove.
9593         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
9594         Change return type to align_flags.
9595         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
9596         Remove not used macro.
9597         (rs6000_loop_align):  Change return type to align_flags.
9598         (rs6000_loop_align_max_skip): Remove.
9599         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
9600         * config/rx/rx-protos.h (rx_align_for_label): Make it
9601         static function.
9602         * config/rx/rx.c (rx_align_for_label): Change return type
9603         to align_flags.
9604         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
9605         macro definitions.
9606         into align_flags class.
9607         (LABEL_ALIGN): Likewise.
9608         (LOOP_ALIGN): Likewise.
9609         * config/s390/s390.c (s390_label_align): Use align_flags
9610         class member.
9611         (s390_asm_output_function_label): Likewise.
9612         * config/sh/sh.c (sh_override_options_after_change):
9613         Use align_flags class directly without macros.
9614         (find_barrier): Likewise.
9615         (barrier_align): Likewise.
9616         (sh_loop_align): Likewise.
9617         * config/spu/spu.c (spu_option_override):
9618         Use align_flags_tuple::get_value instead of removed macros.
9619         (spu_sched_init): Likewise.
9620         * config/spu/spu.h (GTY): Likewise.
9621         * config/visium/visium.c (visium_option_override):
9622         Set "8" as default secondary alignment.
9623         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
9624         in order to guarantee secondary alignment of 8.
9625         * coretypes.h: Include align.h header file.
9626         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
9627         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
9628         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
9629         * doc/tm.texi.in: Likewise.
9630         * final.c (struct label_alignment): Remove not used structure.
9631         (LABEL_ALIGN): Change type to align_flags.
9632         (LOOP_ALIGN): Likewise.
9633         (JUMP_ALIGN): Likewise.
9634         (default_loop_align_max_skip): Remove.
9635         (default_label_align_max_skip): Likewise.
9636         (default_jump_align_max_skip): Likewise.
9637         (default_label_align_after_barrier_max_skip):
9638         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
9639         (LABEL_TO_MAX_SKIP): Remove.
9640         (label_to_alignment): Return align_flags type instead of integer.
9641         (label_to_max_skip): Remove.
9642         (align_fuzz): Use align_flags type.
9643         (compute_alignments): Use align_flags type and use align_flags::max
9644         to combine multiple alignments.
9645         (grow_label_align): Grow vec instead of C array.
9646         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
9647         (shorten_branches):  Use align_flags type and use align_flags::max
9648         to combine multiple alignments.
9649         (final_scan_insn_1): Remove usage of secondary alignment that comes
9650         from label alignment, but instead use proper secondary alignment
9651         which is computed in grow_label_align.
9652         * flags.h (struct align_flags_tuple): Move to align.h.
9653         (struct align_flags): Likewise.
9654         (state_align_loops): Rename to align_loops.
9655         (state_align_jumps): Rename to align_jumps.
9656         (state_align_labels): Rename to align_labels.
9657         (state_align_functions): Rename to align_functions.
9658         (align_loops_log): Remove.
9659         (align_jumps_log): Remove.
9660         (align_labels_log): Remove.
9661         (align_functions_log): Remove.
9662         (align_loops_max_skip): Remove.
9663         (align_jumps_max_skip): Remove.
9664         (align_labels_max_skip): Remove.
9665         (align_functions_max_skip): Remove.
9666         (align_loops_value): Remove.
9667         (align_jumps_value): Remove.
9668         (align_labels_value): Remove.
9669         (align_functions_value): Remove.
9670         * output.h (label_to_alignment): Change return type to align_flags.
9671         (label_to_max_skip): Remove.
9672         * target.def: Remove loop_align_max_skip, label_align_max_skip,
9673         jump_align_max_skip macros.
9674         * targhooks.h (default_loop_align_max_skip): Remove.
9675         (default_label_align_max_skip): Likewise.
9676         (default_jump_align_max_skip): Likewise.
9677         (default_label_align_after_barrier_max_skip): Remove.
9678         * toplev.c (read_log_maxskip): Use ::normalize function.
9679         (parse_N_M): Remove not used argument and also call ::normalize.
9680         (parse_alignment_opts): Do not pass unused arguments.
9681         * varasm.c (assemble_start_function): Use directly align_functions
9682         instead of removed macros.
9683         * system.h: Do not poison removed macros.
9685 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
9687         PR middle-end/86539
9688         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
9689         and cond temporaries don't have reference type if iterator has
9690         pointer type.  For init use &for_pre_body instead of pre_p if
9691         for_pre_body is non-empty.
9693 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9695         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
9696         double-double modes to SFmode directly directly.
9697         (trunc<mode>sf2_fprs): Delete.
9699 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9701         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
9702         for conversions between IFmode and the decimal floating point modes.
9703         (init_float128_ieee): Use the correct names for conversions between
9704         KFmode and the decimal floating point modes.
9706 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
9708         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
9709         for the conversions between TDmode and IFmode.
9710         (init_float128_ieee): Use more correct names for the conversions
9711         between TDmode and KFmode.
9713 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
9715         PR tree-optimization/86526
9716         * builtins.c (expand_builtin_memcmp): Formatting fixes.
9717         (inline_expand_builtin_string_cmp): Likewise.
9718         (inline_string_cmp): Likewise.  Use c_readstr instead of
9719         builtin_memcpy_read_str.  Add unit_mode temporary.
9721 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9723         PR middle-end/86528
9724         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
9725         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
9727 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9729         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
9730         Alphabetize prototypes of built-in functions, separating out
9731         built-in functions that are listed in this section but should be
9732         described elsewhere.
9734 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
9736         PR target/86511
9737         * expmed.c (emit_store_flag): Do not emit setcc followed by a
9738         conditional move when trapping comparison was split to a
9739         non-trapping one (and vice versa).
9741 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
9743         * config/s390/s390.c (s390_function_profiler): Generate nops
9744         instead of profiler call sequences.
9745         * config/s390/s390.opt: Add the new option.
9747 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
9749         * config/s390/s390.c (s390_function_profiler): Generate
9750         __mcount_loc section.
9751         * config/s390/s390.opt: Add the new option.
9753 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
9755         * common.opt: Add the new warning.
9756         * config/s390/s390.c (s390_function_profiler): Emit "brasl
9757         %r0,__fentry__" when -mfentry is specified.
9758         (s390_option_override_internal): Disallow -mfentry for 31-bit
9759         CPUs.
9760         * config/s390/s390.opt: Add the new option.
9762 2018-07-16  Richard Biener  <rguenther@suse.de>
9764         PR lto/86523
9765         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
9766         for function-local FUNCTION_DECL and RESULT_DECL immediately.
9768 2018-07-16  Martin Liska  <mliska@suse.cz>
9770         PR ipa/86529
9771         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
9772         to ::get_create.
9774 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
9776         * config/arc/arcHS.md: Update ARCHS scheduling rules.
9778 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
9780         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
9781         for ARCHS4x.
9782         * config/arc/arc-cpus.def (hs4x): New cpu.
9783         (hs4xd): Likewise.
9784         * config/arc/arc-tables.opt: Regenerate.
9785         * config/arc/arc.c (arc_sched_issue_rate): New function.
9786         (TARGET_SCHED_ISSUE_RATE): Define.
9787         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
9788         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
9789         fpu_cvt.
9790         (attr tune): Add ARCHS4x tune values.
9791         (attr tune_dspmpy): Define.
9792         (*tst): Correct instruction type.
9793         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
9794         * config/arc/arcHS4x.md: New file.
9795         * config/arc/fpu.md: Update instruction type attributes.
9796         * config/arc/t-multilib: Regenerate.
9798 2018-07-16  Tom de Vries  <tdevries@suse.de>
9800         PR debug/86455
9801         * var-tracking.c (vt_initialize): Fix pre_dec handling.
9803 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
9805         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
9806         early clobber.
9808 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
9810         PR tree-optimization/86514
9811         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
9812         conversion to a boolean type from a type with greater precision.
9814 2018-07-16  Tom de Vries  <tdevries@suse.de>
9816         * var-tracking.c (vt_initialize): Print adjusted insn slim if
9817         dump_flags request TDF_SLIM.
9819 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
9821         * fold-const.c (int_const_binop_1): Abstract...
9822         (wide_int_binop): ...wide int code here.
9823         (poly_int_binop): ...poly int code here.
9824         Abstract the rest of int_const_binop_1 into int_const_binop.
9825         * fold-const.h (wide_int_binop): New.
9826         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
9827         Remove useless PLUS/MINUS_EXPR case.
9828         (zero_nonzero_bits_from_vr): Move wide int code...
9829         (zero_nonzero_bits_from_bounds): ...here.
9830         (extract_range_from_binary_expr_1): Move mask optimization code...
9831         (range_easy_mask_min_max): ...here.
9832         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
9833         (range_easy_mask_min_max): New.
9835 2018-07-15  Jeff Law  <law@redhat.com>
9837         PR target/85993
9838         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
9839         block.
9841 2018-07-14  Jim Wilson  <jimw@sifive.com>
9843         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
9845 2018-07-14  Paul Koning  <ni1d@arrl.net>
9847         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
9849 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
9851         * lto-streamer-out.c (copy_function_or_variable): Dump info about
9852         copying section.
9854 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
9855             Steve Munroe  <munroesj52@gmail.com>
9857         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
9858         (_mm_andnot_si128): Likewise.
9859         (_mm_or_si128): Likewise.
9860         (_mm_xor_si128): Likewise.
9862 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
9864         PR middle-end/78809
9865         * builtins.c (expand_builtin_memcmp): Inline the calls first
9866         when result_eq is false.
9867         (expand_builtin_strcmp): Inline the calls first.
9868         (expand_builtin_strncmp): Likewise.
9869         (inline_string_cmp): New routine. Expand a string compare
9870         call by using a sequence of char comparison.
9871         (inline_expand_builtin_string_cmp): New routine. Inline expansion
9872         a call to str(n)cmp/memcmp.
9873         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
9874         New option.
9875         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
9877 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
9879         * config/arm/driver-arm.c: Include arm-native.h.
9880         (host_detect_local_cpu): Use auto-generated data tables.
9881         (vendors, arm_cpu_table): Delete.  Move part information to ...
9882         * config/arm/arm-cpus.in: ... here.
9883         * config/arm/parsecpu.awk (gen_native): New function.
9884         (vendor, part): New CPU fields.
9885         (END): Add support for building the native CPU detection tables.
9886         * config/arm/t-arm (arm-native.h): Add build rule.
9887         (driver-arm.o): Add dependency on arm-native.h.
9889 2018-07-13  Richard Biener  <rguenther@suse.de>
9891         PR middle-end/85974
9892         * match.pd (addr1 - addr2): Allow either of the operand to
9893         have a conversion.
9895 2018-07-13  Tom de Vries  <tdevries@suse.de>
9897         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
9898         in remap_ssa_name.
9900 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
9902         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
9903         arrays instead of numbered variables.
9905 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
9907         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
9908         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
9910 2018-07-13  Richard Biener  <rguenther@suse.de>
9912         PR debug/86452
9913         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
9914         instead of get_context_die.
9916 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
9917             Richard Biener  <rguenther@suse.de>
9919         PR middle-end/86489
9920         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
9921         that the loop latch destination where phi is defined.
9923 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
9925         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
9926         (riscv_expand_epilogue): Add assertion to check interrupt mode.
9927         (riscv_set_current_function): Extract getting interrupt type to new
9928         function.
9929         (riscv_get_interrupt_type): New function.
9930         (riscv_merge_decl_attributes): New function, checking interrupt type is
9931         same.
9932         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
9934 2018-07-12  Paul Koning  <ni1d@arrl.net>
9936         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
9937         directive.
9939 2018-07-12  Paul Koning  <ni1d@arrl.net>
9941         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
9942         zero reference, add doloop_end instead.
9943         * doc/md.texi (decrement_and_branch_until_zero): Remove.
9944         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
9945         detail for doloop_end.
9947 2018-07-12  Martin Sebor  <msebor@redhat.com>
9949         PR c/86453
9950         * attribs.c (decl_attributes): Reject conflicting attributes before
9951         calling attribute handlers.
9953 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
9955         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
9956          parameter.
9957         (gcc::dump_manager::get_dump_file_name): likewise.
9958         (dump_begin): Likewise.
9959         * dumpfile.h (dump_begin): Update prototype.
9960         (gcc::dump_manager::get_dump_file_name,
9961         gcc::dump_manager::get_dump_file_name): Update prototype.
9963 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9965         * internal-fn.h (vectorizable_internal_fn_p): New function.
9966         * tree-vect-slp.c (compatible_calls_p): Likewise.
9967         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
9968         to internal functions.
9969         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
9971 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
9973         * fold-const.h (inverse_conditions_p): Declare.
9974         * fold-const.c (inverse_conditions_p): New function.
9975         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
9976         that test the inverse condition of a conditional internal function.
9977         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
9978         * internal-fn.c (internal_fn_mask_index): Handle conditional
9979         internal functions.
9980         (vectorized_internal_fn_supported_p): New function.
9981         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
9982         (any_pred_load_store): Replace with...
9983         (need_to_predicate): ...this new variable.
9984         (redundant_ssa_names): New variable.
9985         (ifcvt_can_use_mask_load_store): Move initial checks to...
9986         (ifcvt_can_predicate): ...this new function.  Handle tree codes
9987         for which a conditional internal function exists.
9988         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
9989         instead of ifcvt_can_use_mask_load_store.  Update after variable
9990         name change.
9991         (predicate_load_or_store): New function, split out from
9992         predicate_mem_writes.
9993         (check_redundant_cond_expr): New function.
9994         (value_available_p): Likewise.
9995         (predicate_rhs_code): Likewise.
9996         (predicate_mem_writes): Rename to...
9997         (predicate_statements): ...this.  Use predicate_load_or_store
9998         and predicate_rhs_code.
9999         (combine_blocks, tree_if_conversion): Update after above name changes.
10000         (ifcvt_local_dce): Handle redundant_ssa_names.
10001         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
10002         general conditional functions.
10003         * tree-vect-stmts.c (vectorizable_call): Likewise.
10005 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10006             Alan Hayward  <alan.hayward@arm.com>
10007             David Sherwood  <david.sherwood@arm.com>
10009         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
10010         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
10011         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
10012         plus and minus and convert them into IFN_COND_FMA-based sequences.
10013         (convert_mult_to_fma): Handle conditional plus and minus.
10015 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10017         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
10018         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
10019         (cond_fnms_optab): New optabs.
10020         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
10021         internal functions.
10022         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
10023         * internal-fn.h (get_conditional_internal_fn): Declare.
10024         (get_unconditional_internal_fn): Likewise.
10025         * internal-fn.c (cond_ternary_direct): New macro.
10026         (expand_cond_ternary_optab_fn): Likewise.
10027         (direct_cond_ternary_optab_supported_p): Likewise.
10028         (FOR_EACH_COND_FN_PAIR): Likewise.
10029         (get_conditional_internal_fn): New function.
10030         (get_unconditional_internal_fn): Likewise.
10031         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
10032         (gimple_match_op::gimple_match_op): Add a new overload for 5
10033         operands.
10034         (gimple_match_op::set_op): Likewise.
10035         (gimple_resimplify5): Declare.
10036         * genmatch.c (decision_tree::gen): Generate simplifications for
10037         5 operands.
10038         * gimple-match-head.c (gimple_simplify): Define an overload for
10039         5 operands.  Handle calls with 5 arguments in the top-level overload.
10040         (convert_conditional_op): Handle conversions from unconditional
10041         internal functions to conditional ones.
10042         (gimple_resimplify5): New function.
10043         (build_call_internal): Pass a fifth operand.
10044         (maybe_push_res_to_seq): Likewise.
10045         (try_conditional_simplification): Try converting conditional
10046         internal functions to unconditional internal functions.
10047         Handle 3-operand unconditional forms.
10048         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
10049         Define ternary equivalents of the current rules for binary conditional
10050         internal functions.
10051         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
10052         ternary operations.
10053         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
10054         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
10055         (optab): Handle them.
10056         (SVE_COND_FP_TERNARY): New int iterator.
10057         (sve_fmla_op, sve_fmad_op): New int attributes.
10058         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
10059         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
10060         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
10062 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
10064         * target.def (preferred_else_value): New target hook.
10065         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
10066         * doc/tm.texi: Regenerate.
10067         * targhooks.h (default_preferred_else_value): Declare.
10068         * targhooks.c (default_preferred_else_value): New function.
10069         * internal-fn.h (conditional_internal_fn_code): Declare.
10070         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
10071         (get_conditional_internal_fn): Use it.
10072         (conditional_internal_fn_code): New function.
10073         * gimple-match.h (gimple_match_cond): New struct.
10074         (gimple_match_op): Add a cond member function.
10075         (gimple_match_op::gimple_match_op): Update all forms to take a
10076         gimple_match_cond.
10077         * genmatch.c (expr::gen_transform): Use the same condition as res_op
10078         for the suboperation, but don't specify a particular else_value.
10079         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
10080         (visit_nary_op, visit_reference_op_load): Pass
10081         gimple_match_cond::UNCOND to the gimple_match_op constructor.
10082         * gimple-match-head.c: Include tree-eh.h
10083         (convert_conditional_op): New function.
10084         (maybe_resimplify_conditional_op): Likewise.
10085         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
10086         (gimple_resimplify2): Likewise.
10087         (gimple_resimplify3): Likewise.
10088         (gimple_resimplify4): Likewise.
10089         (maybe_push_res_to_seq): Return null for conditional operations.
10090         (try_conditional_simplification): New function.
10091         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
10092         constructor.
10093         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
10094         IFN_COND_* call.
10095         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
10096         function.
10097         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
10099 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
10101         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10102         DECL_FCONTEXT
10103         (hash_tree): Do not hash DECL_FCONTEXT
10104         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
10105         Do not stream DECL_FCONTEXT.
10106         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
10107         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
10109 2018-07-12  Richard Biener  <rguenther@suse.de>
10111         PR debug/86462
10112         * dwarf2out.c (gen_block_die): Only output blocks when they have
10113         at least one !DECL_IGNORED_P variable.
10115 2018-07-12  Richard Biener  <rguenther@suse.de>
10117         PR target/84829
10118         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
10119         Remove -mieee-fp handling.
10121 2018-07-12  Richard Biener  <rguenther@suse.de>
10123         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
10124         left-over from last patch.
10126 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
10128         PR tree-optimization/86492
10129         * gimple-ssa-store-merging.c
10130         (imm_store_chain_info::coalesce_immediate_stores): Call
10131         check_no_overlap even for the merge_overlapping case.  Formatting fix.
10133 2018-07-12  Richard Biener  <rguenther@suse.de>
10135         PR middle-end/86479
10136         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
10137         move possibly trapping operations into the conditional.
10139 2018-07-12  Richard Biener  <rguenther@suse.de>
10141         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
10142         (vn_lookup_simplify_result): Remove recursion limit applied
10143         here.
10144         (vn_nary_build_or_lookup_1): Adjust.
10145         (try_to_simplify): Likewise.
10146         * gimple-match-head.c (gimple_resimplify1): Instead apply one
10147         here.
10148         (gimple_resimplify2): Likewise.
10149         (gimple_resimplify3): Likewise.
10150         (gimple_resimplify4): Likewise.
10152 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
10154         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
10155         Use __mmask64 type instead of __mmask8 for __M argument.
10156         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
10157         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
10158         __U argument.
10159         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
10160         __mmask16 for __M argument.
10161         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
10162         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
10163         to __mmask16 instead of __mmask8.
10164         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
10165         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
10166         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
10167         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
10168         instead of __mmask16 for __U argument.
10169         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
10170         __mmask16 instead of __mmask8 for __U argument.
10171         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
10172         __U argument.
10173         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
10174         __mmask16.
10175         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
10176         argument.
10177         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
10178         __U argument.
10179         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
10180         __mmask16.
10181         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
10182         of __mmask16.
10183         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
10184         __U argument.
10185         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
10186         __U argument.
10187         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
10188         __U argument.
10189         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
10190         __U argument.
10191         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
10192         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
10193         return type as well as __M argument type and all casts from __mmask8
10194         to __mmask32.
10195         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
10196         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
10197         return type as well as __M argument type and all casts from __mmask8
10198         to __mmask16.
10199         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
10200         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
10201         return type as well as __M argument type and all casts from __mmask8
10202         to __mmask32.
10203         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
10204         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
10205         return type as well as __M argument type and all casts from __mmask8
10206         to __mmask16.
10207         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
10208         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
10209         __mmask16.
10211 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
10213         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
10214         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
10215         for __U argument.
10217 2018-07-11  Paul Koning  <ni1d@arrl.net>
10219         * doc/md.texi (define_subst): Document how multiple occurrences of
10220         the same argument in the replacement pattern are handled.
10222 2018-07-11  Paul Koning  <ni1d@arrl.net>
10224         * doc/extend.texi (Common Variable Attributes): Move "mode" into
10225         alphabetical order.
10226         (Common Type Attributes): Add "mode" attribute.
10228 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
10230         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
10231         stream DECL_ORIGINAL_TYPE.
10232         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
10233         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
10234         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
10235         Do not walk original type.
10236         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
10237         external decls.
10238         (write_ts_decl_non_common_tree_pointers): Do not stream
10239         DECL_ORIGINAL_TYPE
10240         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
10241         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
10243 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
10245         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
10246         thread twice from the same starting edge.
10248 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
10250         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
10251         * gimple.c (gimple_call_nonnull_result_p): ...here...
10252         (gimple_call_nonnull_arg): ...and here.
10253         * gimple.h (gimple_call_nonnull_result_p): New.
10254         (gimple_call_nonnull_arg): New.
10256 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
10258         * config/arm/arm-cpus.in: Move information from fpu field of each
10259         cpu definition to the isa field.
10260         * config/arm/parsecpu.awk (fpu): Delete match rule.
10261         (gen_comm_data): Don't add bits from the CPU's FPU entry.
10263 2018-07-11  Richard Biener  <rguenther@suse.de>
10265         PR debug/86457
10266         * dwarf2out.c (init_sections_and_labels): Use
10267         output_asm_line_debug_info consistently.
10268         (dwarf2out_early_finish): Likewise.
10269         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
10270         type units.
10272 2018-07-11  Richard Biener  <rguenther@suse.de>
10274         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
10275         Rework father_bb setting in a way to avoid propagating constants
10276         multiple times on a loop body.
10278 2018-07-10  Mark Wielaard  <mark@klomp.org>
10280         PR debug/86459
10281         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
10283 2018-07-10  Richard Biener  <rguenther@suse.de>
10285         * hash-map.h (hash_map::iterator::operator*): Return
10286         references to key and value.
10288 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
10290         PR c++/86443
10291         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
10292         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
10293         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
10294         (gimplify_omp_for): For composite loops, move outer
10295         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
10296         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
10297         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
10298         TREE_LIST for both the original class iterator and the "last" helper
10299         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
10300         loop, remember has_decl_expr from outer composite loops for the
10301         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
10303 2018-07-09  Martin Sebor  <msebor@redhat.com>
10305         PR middle-end/77357
10306         PR middle-end/86428
10307         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
10308         accessing implicitly initialized array elements.
10309         * expr.c (string_constant): Handle string initializers of
10310         character arrays within aggregates.
10311         * gimple-fold.c (fold_array_ctor_reference): Add argument.
10312         Store element offset.  As a special case, handle zero size.
10313         (fold_nonarray_ctor_reference): Same.
10314         (fold_ctor_reference): Add argument.  Store subobject offset.
10315         * gimple-fold.h (fold_ctor_reference): Add argument.
10317 2018-07-09  Paul Koning  <ni1d@arrl.net>
10319         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
10320         (pdp11_insn_cost): New function.
10321         (pdp11_md_asm_adjust): New function.
10322         (TARGET_INVALID_WITHIN_DOLOOP): Define.
10323         (pdp11_rtx_costs): Update to match machine better.
10324         (output_addr_const_pdp11): Correct format mismatch warnings.
10325         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
10326         * config/pdp11/pdp11.md: General change to add base_cost and/or
10327         length attributes for use by new pdp11_insn_cost function.
10328         (MIN_BRANCH): Correct definition.
10329         (MIN_SOB): Ditto.
10330         (doloop_end): Use standard pattern name for looping pattern.
10331         (doloop_end_nocc): New.
10332         (movsf): Add another constraint alternative.
10333         (zero_extendqihi2): Add constraint alternatives for not in place
10334         extend.
10335         (zero_extendhisi2): Remove.
10336         (shift patterns): Add CC handling variants.
10337         (bswaphi2): New.
10338         (bswapsi2): New.
10339         (rothi3): New.
10340         (define_peephole2): New peephole to recognize mov that sets CC for
10341         subsequent test.
10343 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10345         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
10346         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
10347         wi::add.
10349 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
10351         PR c/86420
10352         * real.c (real_nextafter): Return true if result is denormal.
10354 2018-07-09  Martin Liska  <mliska@suse.cz>
10356         * common.opt: Add back wrongly removed attribute.
10358 2018-07-09  Richard Biener  <rguenther@suse.de>
10360         PR debug/86413
10361         * dwarf2out.c (gen_block_die): For an early generated DIE
10362         always output high/low PC attributes.
10364 2018-07-09  Tom de Vries  <tdevries@suse.de>
10366         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
10367         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
10368         onto VAR_DECL with abstract origin.
10370 2018-07-07  Jim Wilson  <jimw@sifive.com>
10372         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
10374 2018-07-07  Tom de Vries  <tdevries@suse.de>
10376         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
10378 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
10380         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
10381         overflow_type.
10382         (combine_bound): Use wide-int overflow calculation instead of
10383         rolling our own.
10384         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
10385         overflow_type.
10386         * fold-const.c (int_const_binop_2): Same.
10387         (extract_muldiv_1): Same.
10388         (fold_div_compare): Same.
10389         (fold_abs_const): Same.
10390         * match.pd: Same.
10391         * poly-int.h (add): Same.
10392         (sub): Same.
10393         (neg): Same.
10394         (mul): Same.
10395         * predict.c (predict_iv_comparison): Same.
10396         * profile-count.c (slow_safe_scale_64bit): Same.
10397         * simplify-rtx.c (simplify_const_binary_operation): Same.
10398         * tree-chrec.c (tree_fold_binomial): Same.
10399         * tree-data-ref.c (split_constant_offset_1): Same.
10400         * tree-if-conv.c (idx_within_array_bound): Same.
10401         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
10402         * tree-ssa-phiopt.c (minmax_replacement): Same.
10403         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
10404         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
10405         * vr-values.c (vr_values::adjust_range_with_scev): Same.
10406         * wide-int.cc (wi::add_large): Same.
10407         (wi::mul_internal): Same.
10408         (wi::sub_large): Same.
10409         (wi::divmod_internal): Same.
10410         * wide-int.h: Change overflow type to overflow_type for neg, add,
10411         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
10412         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
10413         mul_internal, divmod_internal.
10414         (overflow_type): New enum.
10415         (accumulate_overflow): New.
10417 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
10419         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
10420         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
10422 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
10424         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
10425         argument is checked for zero before entering loop, avoid checking again.
10427 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
10429         * gimplify.h (generic_expr_could_trap_p): Set as global function.
10430         * gimplify.h (generic_expr_could_trap_p): Likwise.
10431         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
10433 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
10435         PR tree-optimization/86401
10436         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
10437         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
10438         (fold_bit_and_mask): ... here.  New helper function for match.pd.
10439         * fold-const.h (fold_bit_and_mask): Declare.
10440         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
10442 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
10444         PR target/86324
10445         * target.def (translate_mode_attribute): New hook.
10446         * targhooks.h (default_translate_mode_attribute): Declare.
10447         * targhooks.c (default_translate_mode_attribute): New function.
10448         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
10449         * doc/tm.texi: Regenerate.
10450         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
10451         (rs6000_translate_mode_attribute): New function.
10453 2018-07-06  Paul Koning  <ni1d@arrl.net>
10455         * doc/md.texi (define_split): Document DONE and FAIL.
10456         (define_peephole2): Ditto.
10458 2018-07-05  Jeff Law  <law@redhat.com>
10460         PR tree-optimization/86010
10461         * tree-ssa-dse.c (compute_trims): More aggressively trim at
10462         both the head and tail of mem* and str* calls.
10464 2018-07-05  Jim Wilson  <jimw@sifive.com>
10466         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
10468 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
10470         * config/aarch64/aarch64-simd.md: correct flags text for
10471         MIN_EXPR replacement.
10473 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
10475         * configure: Regenerated.
10477 2018-07-05  Carl Love  <cel@us.ibm.com>
10479         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
10480         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
10481         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
10482         VSX_BUILTIN_DOUBLEL_V4SF.
10484 2018-07-05  Martin Sebor  <msebor@redhat.com>
10486         PR c++/86400
10487         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
10488         than its domain to compute its the upper bound of a char array.
10490 2018-07-05  Nathan Sidwell  <nathan@acm.org>
10492         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
10493         * doc/cpp.texi: Update comment.
10494         * doc/tm.texi: Rebuilt.
10495         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
10496         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
10497         * doc/extend.texi (Backwards Compatibility): Clarify it is system
10498         headers affected by extern "C".
10499         * system.h: Poison NO_IMPLICIT_EXTERN_C.
10500         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
10501         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
10502         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
10503         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
10504         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
10505         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
10506         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
10507         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
10508         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
10509         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
10510         config/sparc/sp64-elf.h, config/spu/spu.h,
10511         config/stormy16/stormy16.h, config/v850/v850.h,
10512         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
10513         define NO_IMPLICIT_EXTERN_C.
10514         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
10516 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
10518         PR target/84711
10519         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
10520         instead of GET_MODE_SIZE when comparing Units.
10522 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
10524         PR target/84711
10525         * rtlanal.c (set_noop_p): Constrain on mode change,
10526         include hard-reg-set.h
10528 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
10530         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
10532 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
10534         Revert
10535         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
10537         PR sanitizer/84250
10538         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
10539         libasan.
10540         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
10542 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
10544         PR sanitizer/84250
10545         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
10546         libasan.
10547         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
10549 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
10551         PR middle-end/86380
10552         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
10554 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
10556         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
10557         neg_*_op* variables.
10559 2018-07-04  Martin Liska  <mliska@suse.cz>
10561         * tree-switch-conversion.c: Define
10562         max_ratio_for_speed and max_ratio_for_size constants.
10564 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
10565             Martin Liska  <mliska@suse.cz>
10567         PR middle-end/66240
10568         PR target/45996
10569         PR c/84100
10570         * common.opt: Rename align options with 'str_' prefix.
10571         * common/config/i386/i386-common.c (set_malign_value): New
10572         function.
10573         (ix86_handle_option): Use it to set -falign-* options/
10574         * config/aarch64/aarch64-protos.h (struct tune_params): Change
10575         type from int to string.
10576         * config/aarch64/aarch64.c: Update default values from int
10577         to string.
10578         * config/alpha/alpha.c (alpha_override_options_after_change):
10579         Likewise.
10580         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
10581         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10582         max skip conditionally.
10583         * config/i386/freebsd.h (SUBALIGN_LOG): New.
10584         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10585         max skip conditionally.
10586         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10587         max skip conditionally.
10588         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
10589         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10590         max skip conditionally.
10591         * config/i386/i386.c (struct ptt): Change type from int to
10592         string.
10593         (ix86_default_align): Set default values.
10594         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
10595         max skip conditionally.
10596         * config/i386/iamcu.h (SUBALIGN_LOG): New.
10597         (ASM_OUTPUT_MAX_SKIP_ALIGN):
10598         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
10599         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10600         max skip conditionally.
10601         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
10602         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
10603         * config/i386/x86-64.h (SUBALIGN_LOG): New.
10604         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
10605         max skip conditionally.
10606         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10607         * config/ia64/ia64.c (ia64_option_override): Set default values
10608         for alignment options.
10609         * config/m68k/m68k.c: Handle new str_align_* options.
10610         * config/mips/mips.c (mips_set_compression_mode): Change
10611         type of constants.
10612         (mips_option_override): Set default values for options.
10613         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
10614         Likewise.
10615         * config/rs6000/rs6000.c (rs6000_option_override_internal):
10616         Likewise.
10617         * config/rx/rx.c (rx_option_override): Likewise.
10618         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
10619         (LABEL_ALIGN): Use align_labels_log.
10620         (LOOP_ALIGN): Use align_loops_align.
10621         * config/s390/s390.c (s390_asm_output_function_label): Use new
10622         macros.
10623         * config/sh/sh.c (sh_override_options_after_change):
10624         Change type of constants.
10625         * config/spu/spu.c (spu_sched_init): Likewise.
10626         * config/sparc/sparc.c (sparc_option_override): Set default
10627         values for options.
10628         * config/visium/visium.c (visium_option_override): Likewise.
10629         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
10630         emit p2align format with last argument if it's not needed.
10631         * doc/invoke.texi: Document extended format of -falign-*.
10632         * final.c: Use align_labels alignment.
10633         * flags.h (struct target_flag_state): Change type to use
10634         align_flags.
10635         (struct align_flags_tuple): New.
10636         (struct align_flags): Likewise.
10637         (align_loops_log): Redefine macro to use new types.
10638         (align_loops_max_skip): Redefine macro to use new types.
10639         (align_jumps_log): Redefine macro to use new types.
10640         (align_jumps_max_skip): Redefine macro to use new types.
10641         (align_labels_log): Redefine macro to use new types.
10642         (align_labels_max_skip): Redefine macro to use new types.
10643         (align_functions_log): Redefine macro to use new types.
10644         (align_loops): Redefine macro to use new types.
10645         (align_jumps): Redefine macro to use new types.
10646         (align_labels): Redefine macro to use new types.
10647         (align_functions): Redefine macro to use new types.
10648         (align_functions_max_skip): Redefine macro to use new types.
10649         (align_loops_value): New macro.
10650         (align_jumps_value): New macro.
10651         (align_labels_value): New macro.
10652         (align_functions_value): New macro.
10653         * function.c (invoke_set_current_function_hook): Propagate
10654         alignment values from flags to global variables default in
10655         topleev.h.
10656         * ipa-icf.c (sem_function::equals_wpa): Use
10657         cl_optimization_option_eq instead of memcmp.
10658         * lto-streamer.h (cl_optimization_stream_out): Support streaming
10659         of string types.
10660         (cl_optimization_stream_in): Likewise.
10661         * optc-save-gen.awk: Support strings in cl_optimization.
10662         * opth-gen.awk: Likewise.
10663         * opts.c (finish_options): Remove error checking of invalid
10664         value ranges.
10665         (MAX_CODE_ALIGN): Remove.
10666         (MAX_CODE_ALIGN_VALUE): Likewise.
10667         (parse_and_check_align_values): New function.
10668         (check_alignment_argument): Likewise.
10669         (common_handle_option): Use check_alignment_argument.
10670         * opts.h (parse_and_check_align_values): Declare.
10671         * toplev.c (init_alignments): Remove.
10672         (read_log_maxskip): New.
10673         (parse_N_M): Likewise.
10674         (parse_alignment_opts): Likewise.
10675         (backend_init_target): Remove usage of init_alignments.
10676         * toplev.h (parse_alignment_opts): Declare.
10677         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
10678         argument.
10679         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
10680         * tree.c (cl_option_hasher::equal): New.
10681         * varasm.c: Use new global macros.
10683 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
10685         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
10686         Use a simpler align directive also if MAXSKIP = ALIGN-1.
10687         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10688         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10689         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10690         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
10691         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
10692         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
10693         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
10694         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10695         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10696         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10697         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
10699 2018-07-04  Martin Liska  <mliska@suse.cz>
10700             Jonathan Wakely  <jwakely@redhat.com>
10702         * coverage.c: Use correct type.
10703         * doc/invoke.texi: Language correction.
10705 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
10707         PR target/85620
10708         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
10709         ENDBRANCH for non-tail call which may return via indirect branch.
10710         * doc/extend.texi: Document indirect_return attribute.
10712 2018-07-03  Martin Sebor  <msebor@redhat.com>
10714         PR tree-optimization/86274
10715         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
10716         precondition.
10717         (format_floating): Correct handling of infinities and NaNs.
10719 2018-07-03  Martin Sebor  <msebor@redhat.com>
10721         * print-tree.c (print_real_cst): New function.
10722         (print_node_brief): Call it.
10723         (print_node): Ditto.
10725 2018-07-03  Jeff Law  <law@redhat.com>
10727         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
10728         into a single pattern.
10730         * config/h8300/h8300.md (ors code_iterator): New.
10731         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
10732         a single pattern and single splitter.
10733         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
10734         (iorqi3_1, xorqi3_1): Likewise.
10735         (iorqi3, xorqi3 expanders): Similarly.
10737         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
10738         (movmd_internal) into a single pattern using the P mode iterator.
10739         (movmd splitters): Similarly.
10740         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
10741         (movsd splitters): Similarly.
10743         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
10744         ADDB, ADDW and ADDL into a single ADD attribute which selects the
10745         right table based on the size of the operand.
10746         * config/h8300/h8300.md (length_table): Corresponding changes. All
10747         references to "addb", "addw" and "addl" changed to "add".
10748         (btst patterns): Merge two variants into a single pattern.
10749         (tstqi, tsthi): Likewise.
10750         (addhi3_incdec, addsi3_incdec): Likewise.
10751         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
10752         (mulhi3, mulsi3): Likewise.
10753         (udivhi3, udivsi3): Likewise.
10754         (divhi3, divsi3): Likewise.
10755         (andorqi3, andorhi3, andorsi3): Likewise.
10757 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
10759         PR target/85694
10760         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
10761         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
10763 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10765         PR tree-optimization/85694
10766         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
10767         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
10768         UNSPEC_URHADD.
10769         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
10770         (<u>avg<mode>3_ceil): New patterns.
10772 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
10774         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
10775         scan-tree-dump directive.
10776         * gcc.dg/vect/slp-perm-2.c: Likewise.
10777         * gcc.dg/vect/slp-perm-3.c: Likewise.
10778         * gcc.dg/vect/slp-perm-5.c: Likewise.
10779         * gcc.dg/vect/slp-perm-6.c: Likewise.
10780         * gcc.dg/vect/slp-perm-7.c: Likewise.
10781         * gcc.dg/vect/slp-perm-8.c: Likewise.
10783 2018-07-03  Marek Polacek  <polacek@redhat.com>
10785         PR middle-end/86202
10786         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
10788 2018-07-03  Richard Biener  <rguenther@suse.de>
10790         PR ipa/86389
10791         * tree-ssa-structalias.c (find_func_clobbers): Properly
10792         handle indirect calls.
10794 2018-07-03  Jeff Law  <law@redhat.com>
10796         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
10797         (shifts): New code iterator.
10798         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
10799         expander.  Fix HImode handling on H8/SX.
10800         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
10801         (subqi3, subhi3, subsi3 expanders): Likewise.
10802         (andqi3, andhi3, andsi3 expanders): Likewise.
10803         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
10804         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
10805         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
10806         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
10807         (zero_extendqihi2, zero_extendqisi2): Likewise.
10808         (extendqihi2, extendqisi2): Likewise.
10809         (rotlqi3, rotlhi3, rotlsi3): Likewise.
10810         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
10811         (rotlqi3_1, rotlhi3_1): Likewise.
10812         (logicalhi3_sn, logicalsi3_sn): Likewise.
10813         (logicalhi3, logicalsi3): Likewise.
10815 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10817         * tree-vect-patterns.c (vect_recog_rotate_pattern)
10818         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
10819         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
10820         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
10821         type to append_pattern_def_seq instead of creating a stmt_vec_info
10822         directly.
10823         (build_mask_conversion): Likewise.  Remove vinfo argument.
10824         (vect_add_conversion_to_patterm): Likewise, renaming to...
10825         (vect_add_conversion_to_pattern): ...this.
10826         (vect_recog_mask_conversion_pattern): Update call to
10827         build_mask_conversion.  Pass the vector type to
10828         append_pattern_def_seq here too.
10829         (vect_recog_gather_scatter_pattern): Update call to
10830         vect_add_conversion_to_pattern.
10832 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10834         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
10835         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10836         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
10837         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
10838         STMT_VINFO_PATTERN_DEF_SEQ to null here.
10839         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
10840         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
10841         append_pattern_def_seq instead of new_pattern_def_seq.
10842         (vect_recog_divmod_pattern): Do both of the above.
10843         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
10844         is null.
10846 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10848         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
10849         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
10850         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
10851         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
10852         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
10853         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
10854         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
10855         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10856         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10857         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
10858         parameter with a single stmt_vec_info.
10859         (vect_recog_func_ptr): Likewise.
10860         (vect_recog_gather_scatter_pattern): Likewise, folding in...
10861         (vect_try_gather_scatter_pattern): ...this.
10862         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
10863         the stmt_vec_info of the statement to be matched.  Don't clear
10864         STMT_VINFO_RELATED_STMT.
10865         (vect_pattern_recog): Update call accordingly.
10867 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10869         PR tree-optimization/85694
10870         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
10871         (uavgM3_ceil): Document new optabs.
10872         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
10873         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
10874         functions.
10875         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
10876         (savg_ceil_optab): New optabs.
10877         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
10878         (vect_vect_recog_func_ptrs): Add it.
10879         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
10880         constant directly from the associated lhs.
10882 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10884         * tree-vect-patterns.c (vect_split_statement): New function.
10885         (vect_convert_input): Use it to try to split an existing cast.
10887 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10889         * poly-int.h (print_hex): New function.
10890         * dumpfile.h (dump_dec, dump_hex): Declare.
10891         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
10892         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
10893         min_input_precision, operation_precision and operation_sign.
10894         * tree-vect-patterns.c (vect_get_range_info): New function.
10895         (vect_same_loop_or_bb_p, vect_single_imm_use)
10896         (vect_operation_fits_smaller_type): Delete.
10897         (vect_look_through_possible_promotion): Add an optional
10898         single_use_p parameter.
10899         (vect_recog_over_widening_pattern): Rewrite to use new
10900         stmt_vec_info infomration.  Handle one operation at a time.
10901         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
10902         (vect_truncatable_operation_p, vect_set_operation_type)
10903         (vect_set_min_input_precision): New functions.
10904         (vect_determine_min_output_precision_1): Likewise.
10905         (vect_determine_min_output_precision): Likewise.
10906         (vect_determine_precisions_from_range): Likewise.
10907         (vect_determine_precisions_from_users): Likewise.
10908         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
10909         (vect_vect_recog_func_ptrs): Put over_widening first.
10910         Add cast_forwprop.
10911         (vect_pattern_recog): Call vect_determine_precisions.
10913 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
10915         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
10916         statements that have been replaced by further pattern statements.
10917         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
10919 2018-07-03  Richard Biener  <rguenther@suse.de>
10921         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
10922         always set *dt.  Dump vectype in vectype overload.
10923         * dumpfile.h (dump_gimple_expr): New function.
10924         (dump_gimple_expr_loc): Likewise.
10925         * dumpfile.c (dump_gimple_expr): New function.
10926         (dump_gimple_expr_loc): Likewise.
10928 2018-07-02  Jeff Law  <law@redhat.com>
10930         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
10931         the H8/300, H8/300H and H8/S variants into a single pattern.
10932         (movhi_h8300, movqi_h8300hs): Similarly.
10933         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
10934         (QHI mode iterator): New.
10936         * config/h8300/h8300.md: Remove trailing whitespace.
10938 2018-07-02  Jim Wilson  <jimw@sifive.com>
10940         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
10941         instead of emit_insn for interrupt returns.
10942         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
10943         (riscv_sret, riscv_uret): Likewise.
10945 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
10947         * pretty-print.c (selftest::test_pp_format): Move save and restore
10948         of quotes to class auto_fix_quotes, and add an instance.
10949         * selftest.c: Include "intl.h".
10950         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
10951         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
10952         * selftest.h (selftest::auto_fix_quotes): New class.
10954 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
10956         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
10957         (aarch64_sve_prepare_conditional_op): Remove.
10958         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
10959         Allow aarch64_simd_reg_or_zero as select operand; remove
10960         the aarch64_sve_prepare_conditional_op call.
10961         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
10962         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
10963         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
10964         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
10965         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
10966         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
10967         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
10968         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
10969         and a splitters to match all of the *_any patterns.
10970         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
10972         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
10973         (SVE_COND_FP_BINARY_REV): Remove.
10974         (sve_int_op_rev, sve_fp_op_rev): New.
10975         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
10976         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
10977         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
10978         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
10979         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
10980         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
10981         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
10982         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
10983         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
10985         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
10986         Remove match_dup 1 from the inner unspec.
10987         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
10989         * config/aarch64/aarch64.md (movprfx): New attr.
10990         (length): Default movprfx to 8.
10991         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
10992         (*madd<SVE_I>, *msub<SVE_I): Likewise.
10993         (*<su>mul<SVE_I>3_highpart): Likewise.
10994         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
10995         (*v<ASHIFT><SVE_I>3): Likewise.
10996         (*<su><MAXMIN><SVE_I>3): Likewise.
10997         (*<su><MAXMIN><SVE_F>3): Likewise.
10998         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
10999         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
11000         (*div<SVE_F>4): Likewise.
11002 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
11004         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
11005         in dump string.
11007 2018-07-02  Richard Biener  <rguenther@suse.de>
11009         PR tree-optimization/86363
11010         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
11011         memset argument refers to a non-variable address.
11013 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
11015         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
11016         {PLUS,MINUS}_EXPR code to...
11017         (adjust_symbolic_bound): ...here,
11018         (combine_bound): ...here,
11019         (set_value_range_with_overflow): ...and here.
11021 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
11023         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
11024         code...
11025         (extract_range_from_abs_expr): ...here.
11027 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
11029         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
11030         -fno-omit-frame-pointer when not optimizing.
11032 2018-07-02  Martin Liska  <mliska@suse.cz>
11034         PR ipa/86279
11035         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
11036         (propagate_nothrow): Likewise.
11038 2018-07-02  Martin Liska  <mliska@suse.cz>
11040         PR ipa/86323
11041         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
11043 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
11045         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
11046         function in r262149, changing "loc" param from source_location to
11047         const dump_location_t &.
11048         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
11049         declaration, as above.
11051 2018-07-01  Paul Koning  <ni1d@arrl.net>
11053         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
11054         -munit-asm, -mgnu-asm, -mdec-asm.
11055         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
11056         (pdp11_output_labelref): New.
11057         (pdp11_output_def): New.
11058         (pdp11_output_addr_vec_elt): New.
11059         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
11060         %# and %@ format codes.
11061         (pdp11_option_override): New.
11062         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
11063         (pdp11_output_ident): New.
11064         (pdp11_asm_named_section): New.
11065         (pdp11_asm_init_sections): New.
11066         (pdp11_file_start): New.
11067         (pdp11_file_end): New.
11068         (output_ascii): Use .ascii/.asciz for -mdec-asm.
11069         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
11070         %o, like %c but octal.
11071         (pdp11_option_override): New.
11072         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
11073         -mdec-asm.
11074         (DATA_SECTION_ASM_OP): Ditto.
11075         (READONLY_DATA_SECTION_ASM_OP): New.
11076         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
11077         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
11078         (ASM_OUTPUT_LABELREF): Ditto.
11079         (ASM_OUTPUT_DEF): Ditto.
11080         (ASM_OUTPUT_EXTERNAL): New.
11081         (ASM_OUTPUT_SOURCE_FILENAME): New.
11082         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
11083         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
11084         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
11085         %# and %@ format codes.
11086         * config/pdp11/pdp11.opt (mgnu-asm): New.
11087         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
11088         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
11089         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
11091 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
11093         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
11094         dereferencing path[] beyond its length.
11095         (debug_path): New.
11096         (debug_all_paths): New.
11097         (rewire_first_differing_edge): New.
11098         (adjust_paths_after_duplication): New.
11099         (duplicate_thread_path): Call adjust_paths_after_duplication.
11100         Add new argument.
11101         (thread_through_all_blocks): Add new argument to
11102         duplicate_thread_path.
11104 2018-06-30  Jim Wilson  <jimw@sifive.com>
11106         * config/riscv/predicates.md (p2m1_shift_operand): New.
11107         (high_mask_shift_operand): New.
11108         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
11109         pattern using p2m1_shift_operand.
11110         (lshsi3_zero_extend_3+2): New combiner pattern using
11111         high_mask_shift_operand.
11113 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11115         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
11116         split out from...
11117         (vect_recog_rotate_pattern): ...here.
11118         (vect_convert_input): Try to insert casts of invariants in the
11119         preheader.
11120         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
11121         preheader to be empty.
11123 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11125         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
11126         vector type.  If given, install it in the new statement's
11127         STMT_VINFO_VECTYPE.
11128         (vect_element_precision): New function.
11129         (vect_unpromoted_value): New struct.
11130         (vect_unpromoted_value::vect_unpromoted_value): New function.
11131         (vect_unpromoted_value::set_op): Likewise.
11132         (vect_look_through_possible_promotion): Likewise.
11133         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
11134         (vect_widened_op_tree, vect_convert_input): Likewise.
11135         (vect_convert_inputs, vect_convert_output): Likewise.
11136         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
11137         to handle the optional cast of the multiplication result and
11138         vect_widened_op_tree to detect the widened multiplication itself.
11139         Do not require the input and output of promotion casts to have
11140         the same sign, but base the signedness of the operation on the
11141         input rather than the result.  If the pattern includes two
11142         promotions, check that those promotions have the same sign.
11143         Do not restrict the MULT_EXPR handling to a double-width result;
11144         handle quadruple-width results and wider.  Use vect_convert_inputs
11145         to convert the inputs to the common type.
11146         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
11147         to handle the optional cast of the ABS result.  Also allow a sign
11148         change or a sign extension between the ABS and MINUS.
11149         Use vect_widened_op_tree to detect the widened subtraction and use
11150         vect_convert_inputs to convert the inputs to the common type.
11151         (vect_handle_widen_op_by_const): Delete.
11152         (vect_recog_widen_op_pattern): New function.
11153         (vect_recog_widen_mult_pattern): Use it.
11154         (vect_recog_widen_shift_pattern): Likewise.
11155         (vect_recog_widen_sum_pattern): Use
11156         vect_look_through_possible_promotion to handle the promoted
11157         PLUS_EXPR operand.
11159 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11161         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
11162         the containing gimple_seq *.
11163         * gimple-iterator.h (gsi_for_stmt): Declare it.
11164         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
11165         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
11166         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
11167         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
11168         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
11169         checks.
11170         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
11171         split out from...
11172         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
11173         statement being replaced is part of an existing pattern
11174         definition sequence, inserting the new pattern statements before
11175         the original one.
11176         (vect_pattern_recog_1): Don't return a bool.  If the statement
11177         is already part of a pattern, instead apply pattern matching
11178         to the pattern definition statements.  Don't clear the
11179         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
11180         (vect_pattern_recog): Don't break after the first match;
11181         continue processing the pattern definition statements instead.
11182         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
11184 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11186         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11187         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
11188         (vect_recog_widen_sum_pattern): Use it.
11190 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11192         * tree-vect-loop.c (vectorizable_reduction): Assert that the
11193         phi is not a pattern statement and has not been replaced by
11194         a pattern statement.
11195         * tree-vect-patterns.c (type_conversion_p): Don't check
11196         STMT_VINFO_IN_PATTERN_P.
11197         (vect_recog_vector_vector_shift_pattern): Likewise.
11198         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
11199         the pattern statement rather than the original statement; check
11200         directly for a WIDEN_MULT_EXPR here.
11201         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
11202         vect_is_simple_use to return the pattern statement rather
11203         than the original statement; use is_pattern_stmt_p to check
11204         for such a pattern statement.
11205         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
11206         to return the pattern statement rather than the original statement;
11207         don't do the same transformation here.
11208         (vect_is_simple_use): If the defining statement has been replaced
11209         by a pattern statement, return the pattern statement instead.
11210         Remove the corresponding (local) transformation from the vectype
11211         overload.
11213 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
11215         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
11216         end and default to null.
11217         * tree-vect-loop.c (vect_create_epilog_for_reduction)
11218         (vectorizable_reduction): Update calls accordingly, dropping the
11219         gimple ** argument if the passed-back statement isn't needed.
11220         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
11221         (vect_recog_rotate_pattern): Likewise.
11222         (vect_recog_mask_conversion_pattern): Likewise.
11223         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11224         (vect_mask_constant_operand_p): Likewise.
11225         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
11226         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
11227         (get_group_load_store_type, get_load_store_type): Likewise.
11228         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
11229         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
11230         (vectorizable_conversion, vectorizable_assignment): Likewise.
11231         (vectorizable_shift, vectorizable_operation): Likewise.
11232         (vectorizable_store, vect_is_simple_cond): Likewise.
11233         (vectorizable_condition, vectorizable_comparison): Likewise.
11234         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
11235         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
11236         and move it to the end.  Cope with null def_stmt_outs.
11238 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11240         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
11242 2018-06-29  Jeff Law  <law@redhat.com>
11244         * config/v850/v850.c (v850_legitimate_address_p): Handle large
11245         displacements for TARGET_V850E2V3 and newer.
11246         (TARGET_LRA_P): Remove.  Defaults to LRA now.
11247         * config/v850/v850.md (sign23byte_load): Remove.
11248         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
11249         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
11251 2018-06-29  Martin Liska  <mliska@suse.cz>
11253         PR lto/85759
11254         * coverage.c (coverage_init): Mangle full path name.
11255         * doc/invoke.texi: Document the change.
11256         * gcov-io.c (mangle_path): New.
11257         * gcov-io.h (mangle_path): Likewise.
11258         * gcov.c (mangle_name): Use mangle_path for path mangling.
11260 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11262         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
11263         if starting source register is not even.
11265 2018-06-29  Martin Liska  <mliska@suse.cz>
11267         PR tree-optimization/86263
11268         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
11269         Make edge redirection.
11271 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
11273         * dumpfile.c (dump_loc): Add indentation based on scope depth.
11274         (dump_scope_depth): New variable.
11275         (get_dump_scope_depth): New function.
11276         (dump_begin_scope): New function.
11277         (dump_end_scope): New function.
11278         * dumpfile.h (get_dump_scope_depth): New declaration.
11279         (dump_begin_scope): New declaration.
11280         (dump_end_scope): New declaration.
11281         (class auto_dump_scope): New class.
11282         (AUTO_DUMP_SCOPE): New macro.
11283         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
11284         AUTO_DUMP_SCOPE.
11286 2018-06-29  Richard Biener  <rguenther@suse.de>
11288         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
11289         compute_all_dependences succeeds.
11290         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
11291         exceed --param loop-max-datarefs-for-datadeps.
11293 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
11295         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
11297 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
11299         PR target/86348
11300         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
11301         alternative 0 in preferred_for_speed attribute.
11303 2018-06-28  Paul Koning  <ni1d@arrl.net>
11305         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
11306         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
11307         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
11308         * config/pdp11/pdp11.md: Correct "length" attribute calculation
11309         for shift insn patterns.
11311 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
11313         * cgraph.c (cgraph_node::get_body): Replace assignments to
11314         "dump_file" with calls to set_dump_file.
11315         * dumpfile.c (alt_dump_file): Make static, and group with...
11316         (alt_flags): ...this definition.
11317         (dumps_are_enabled): New variable.
11318         (refresh_dumps_are_enabled): New function.
11319         (set_dump_file): New function.
11320         (set_alt_dump_file): New function.
11321         (gcc::dump_manager::dump_start): Replace assignments to
11322         "dump_file" and "alt_dump_file" with calls to set_dump_file and
11323         set_alt_dump_file.
11324         (gcc::dump_manager::dump_finish): Likewise.
11325         * dumpfile.h (alt_dump_file): Delete decl.
11326         (dumps_are_enabled): New variable decl.
11327         (set_dump_file): New function decl.
11328         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
11329         global.
11330         * tree-nested.c (lower_nested_functions): Replace assignments to
11331         "dump_file" with calls to set_dump_file.
11333 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
11335         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
11336         goto_locus of each outgoing edge of each basic block.
11338 2018-06-28  Richard Biener  <rguenther@suse.de>
11340         * dwarf2out.c (decl_scope_table): Remove.
11341         (push_decl_scope): Likewise.
11342         (pop_decl_scope): Likewise.
11343         (gen_type_die_for_member): Do not call push/pop_decl_scope.
11344         (gen_struct_or_union_type_die): Likewise.
11345         (gen_tagged_type_die): Likewise.
11346         (dwarf2out_init): Do not initialize decl_scope_table.
11347         (dwarf2out_c_finalize): Do not free it.
11349 2018-06-28  Richard Biener  <rguenther@suse.de>
11351         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
11352         deciding whether to not re-use a DIE.
11354 2018-06-28  Richard Biener  <rguenther@suse.de>
11356         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
11357         DW_AT_abstract_origin attribute.
11359 2018-06-28  Martin Liska  <mliska@suse.cz>
11361         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
11362         Use newly introduced constants.
11363         * tree-switch-conversion.h (struct jump_table_cluster):
11364         Define max_ratio_for_size and max_ratio_for_speed.
11366 2018-06-28  Martin Liska  <mliska@suse.cz>
11368         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
11369         Add new checking assert to catch invalid state.
11370         (jump_table_cluster::can_be_handled): Handle single case
11371         clusters.
11372         (jump_table_cluster::is_beneficial): Bail out for such case.
11373         (bit_test_cluster::find_bit_tests):
11374         Add new checking assert to catch invalid state.
11375         (bit_test_cluster::can_be_handled): Handle single case
11376         clusters.
11377         (bit_test_cluster::is_beneficial): Bail out for such case.
11378         (switch_decision_tree::analyze_switch_statement):
11379         Fix comment.
11381 2018-06-28  Martin Liska  <mliska@suse.cz>
11383         * common.opt: Introduce -completion option.
11384         * gcc.c (driver_handle_option): Handle it.
11385         (driver::main): Print completions if completion
11386         is set.
11387         * opt-suggestions.c (option_proposer::get_completions):
11388         New function.
11389         (option_proposer::suggest_completion): Likewise.
11390         (option_proposer::find_param_completions): Likewise.
11391         (verify_autocompletions): Likewise.
11392         (test_completion_valid_options): Likewise.
11393         (test_completion_valid_params): Likewise.
11394         (in_completion_p): Likewise.
11395         (empty_completion_p): Likewise.
11396         (test_completion_partial_match): Likewise.
11397         (test_completion_garbage): Likewise.
11398         (opt_proposer_c_tests): Likewise.
11399         * opt-suggestions.h: Declare new functions.
11400         * opts.c (common_handle_option): Handle OPT__completion_.
11401         * selftest-run-tests.c (selftest::run_tests): Add
11402         opt_proposer_c_tests.
11403         * selftest.c (assert_str_startswith): New.
11404         * selftest.h (assert_str_startswith): Likewise.
11405         (opt_proposer_c_tests): New.
11406         (ASSERT_STR_STARTSWITH): Likewise.
11408 2018-06-28  Martin Liska  <mliska@suse.cz>
11410         * Makefile.in: Add opt-suggestions.o.
11411         * gcc-main.c: Include opt-suggestions.h.
11412         * gcc.c (driver::driver): Likewise.
11413         (driver::~driver): Remove m_option_suggestions.
11414         (driver::build_option_suggestions): Moved to option_proposer.
11415         (driver::suggest_option): Likewise.
11416         (driver::handle_unrecognized_options): Use option_proposer.
11417         * gcc.h (class driver): Add new memver m_option_proposer.
11418         * opt-suggestions.c: New file.
11419         * opt-suggestions.h: New file.
11421 2018-06-28  Martin Liska  <mliska@suse.cz>
11423         * vec.h (class auto_string_vec): New (moved from auto_argvec).
11424         (auto_string_vec::~auto_string_vec): Likewise.
11426 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
11428         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
11429         prevent_decl_creation_for_types fields up and add reset_location field.
11430         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
11431         statement if id->reset_location is true.
11432         (copy_edges_for_bb): Do not set goto_locus on the new edges if
11433         id->reset_location is true.
11434         (copy_phis_for_bb): Force input_location on the arguments if
11435         id->reset_location is true.
11436         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
11437         is set on the function to be inlined.
11439 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
11441         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
11443 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
11445         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
11446         registers for Pmode.
11447         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
11448         hard registers for the clobbered pseudo.
11450 2018-06-27  Paul Koning  <ni1d@arrl.net>
11452         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
11453         mutually exclusive options.
11454         * config/pdp11/constraints.md (h): New constraint.
11455         (O): Update definition to match shift code generation.
11456         (D): New constraint.
11457         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
11458         (CCFP): Remove.
11459         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
11460         function.
11461         (output_jump): Change arguments.
11462         (pdp11_fixed_cc_regs): New function.
11463         (pdp11_cc_mode): Ditto.
11464         (pdp11_expand_shift): Ditto.
11465         (pdp11_assemble_shift): Ditto.
11466         (pdp11_small_shift): Ditto.
11467         (pdp11_branch_cost): Remove.
11468         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
11469         from output.
11470         (pdp11_register_move_cost): Update for CC registers.
11471         (pdp11_rtx_costs): Add case for LSHIFTRT.
11472         (pdp11_output_jump): Add CCNZ mode conditional branches.
11473         (notice_update_cc_on_set): Remove.
11474         (pdp11_cc_mode): New function.
11475         (simple_memory_operand): Correct pre/post decrement case.
11476         (no_side_effect_operand): New function.
11477         (pdp11_regno_reg_class): Add CC_REGS class.
11478         (pdp11_fixed_cc_regs): New function.
11479         (pdp11_small_shift): New function.
11480         (pdp11_expand_shift): New function to expand shift insns.
11481         (pdp11_assemble_shift): New function to output shifts.
11482         (pdp11_branch_cost): Remove.
11483         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
11484         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
11485         (WCHAR_TYPE): Ditto.
11486         (PTRDIFF_TYPE): Ditto.
11487         (ADJUST_INSN_LENGTH): New macro.
11488         (FIXED_REGISTERS): Add CC registers.
11489         (CALL_USED_REGISTERS): Ditto.
11490         (reg_class): Ditto.
11491         (REG_CLASS_NAMES): Ditto.
11492         (REG_CLASS_CONTENTS): Ditto.
11493         (SELECT_CC_MODE): Use new function.
11494         (TARGET_FLAGS_REGNUM): New macro.
11495         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
11496         (cc0_reg_rtx): Remove.
11497         (CC_STATUS_MDEP): Remove.
11498         (CC_STATUS_MDEFP_INIT): Remove.
11499         (CC_IN_FPU): Remove.
11500         (NOTICE_UPDATE_CC): Remove.
11501         (REGISTER_NAMES): Add CC registers.
11502         (BRANCH_COST): Change to constant 1.
11503         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
11504         handling.
11505         * config/pdp11/pdp11.opt (mbcopy): Remove.
11506         (mbcopy-builtin): Remove.
11507         (mbranch-cheap): Remove.
11508         (mbranch-expensive): Remove.
11509         * config/pdp11/predicates.md (expand_shift_operand): Update to
11510         match shift code generation.
11511         (ccnz_operator): New predicate.
11512         * doc/invoke.texi (PDP-11 Options): Remove deleted options
11513         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
11514         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
11515         exclusive options.
11516         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
11517         description of O constraint.
11519 2018-06-27  Jeff Law  <law@redhat.com>
11520             Austin Law  <austinklaw@gmail.com>
11522         * config/v850/v850.md (addsi3_set_flags): New pattern.
11523         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
11524         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
11525         (zero_extendhisi2_v850_set_flags): Likewise.
11526         (zero_extendqisi2_v850_set_flags): Likewise.
11527         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
11528         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
11529         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
11531         * config/v850/v850-protos.h (notice_update_cc): Remove.
11532         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
11533         (v850_print_operand): Handle 'D' and "d".
11534         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
11535         Add handling of arithmetic/logical operations compared against zero.
11536         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
11537         Do not look at v850_compare_op, instead get mode from last argument.
11538         (v850_gen_compare): Remove
11539         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
11540         after reload for prologue insns.
11541         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
11542         patterns.
11543         (construct_save_jarl): Likewise.
11544         (TARGET_FLAGS_REGNUM): Define.
11545         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
11546         (NOTICE_UPDATE_CC): Remove.
11547         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
11548         than cc0.  Conditionalize on reload_completed.
11549         (cmpsi_insn, setfcc_insn): Likewise.
11550         (tst1 splitter): Turn into define_and_split which sets the flags
11551         after reload.
11552         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
11553         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
11554         (cstoresf4, cstoredf4): Clobber the flags.
11555         (cmpsi, cmpsf, cmpdf): Remove expanders.
11556         (setf_insn): Remove pattern.
11557         (addsi3): Turn into define_and_split which clobbers the flags after
11558         reload and a suitable pattern (addsi3_clobber_flags) for use after
11559         reload.
11560         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
11561         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
11562         (ashrsi3, ashrsi3_v850e2): Likewise.
11563         (bins): Clobber the flags.
11564         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
11565         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
11566         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
11567         (call_value_internal_short, call_value_internal_long): Likewise.
11568         (callt_save_interrupt, callt_return_interrupt): Likewise.
11569         (save_interrupt, return_interrupt): Likewise.
11570         (callt_save_all_interrupt, save_all_interrupt): Likewise.
11571         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
11572         (restore_all_interrupt, _restore_all_interrupt): Likewise.
11573         (All FP comparisons): Only allow after reload has completed.
11574         (trfsr): Likewise.
11575         (divh, divhu): Tweak output template.
11576         (branch_z_normal, branch_z_invert): Remove
11577         (branch_nz_normal, branch_nz_invert): Likewise.
11578         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
11580         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
11581         * config/v850/v850.c (notice_update_cc): Remove.
11582         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
11583         (CC_NO_CARRY): Likewise.
11584         (NOTICE_UPDATE_CC): Define to nothing.
11585         * config/v850/v850.md: Remove block comment on cc0 handling
11586         Remove "cc" attribute from all patterns.  Remove cc_status handling
11587         from all patterns.  Minor formatting fixes.
11589 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11591         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
11592         (cortex-a76.cortex-a55): Likewise.
11593         * config/aarch64/aarch64-tune.md: Regenerate.
11594         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
11595         cortex-a76.cortex-a55.
11597 2018-06-27  Jeff Law  <law@redhat.com>
11599         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
11600         (MULTILIB_DIRNAMES): Similarly.
11602 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
11604         * gimple.h (gimple_return_retbnd): Delete.
11605         (gimple_return_set_retbnd): Likewise.
11606         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
11607         gimple_return_set_retbnd.
11608         * gimple-pretty-print.c (dump_gimple_return): Remove call to
11609         gimple_return_retbnd and adjust.
11610         * tree-inline.h (struct copy_body_data): Remove retbnd field.
11611         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
11612         Explicitly return NULL in a couple more cases.  Move assertion
11613         on debug statements and remove unreachable code.
11614         (reset_debug_binding): Do not test id->retbnd.
11615         (expand_call_inline): Do not set it.
11617 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
11619         * configure.ac: Add --disable-gcov option.
11620         * configure: Regenerate.
11621         * Makefile.in: Honour @enable_gcov@.
11622         * doc/install.texi: Document --disable-gcov.
11624 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11626         * config/arm/arm-cpus.in (cortex-a76): New entry.
11627         (cortex-a76.cortex-a55): Likewise.
11628         * config/arm/arm-tables.opt: Regenerate.
11629         * config/arm/arm-tune.md: Likewise.
11630         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
11631         * doc/invoke.texi (ARM Options): Document cortex-a76 and
11632         cortex-a76.cortex-a55.
11634 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
11636         PR target/85769
11637         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
11639 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
11641         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
11642         comment.
11643         (EPILOGUE_USES): Likewise.
11645 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
11647         * tree-inline.c (remap_location): New function extracted from...
11648         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
11649         (copy_phis_for_bb): ...here.  Call remap_location.
11650         (copy_cfg_body): Adjust call to copy_edges_for_bb.
11652 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
11654         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
11655         unaligned vsx for 16B memset.
11657 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
11659         PR target/86285
11660         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
11661         ieee128_float_type_node to long_double_type_node unless
11662         TARGET_LONG_DOUBLE_128 is set.
11664 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
11666         * cfgloop.c (get_loop_location): Convert return type from
11667         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
11668         by implicit construction from rtx_insn *, and using
11669         dump_user_location_t::from_function_decl for the fallback case.
11670         * cfgloop.h (get_loop_location): Convert return type from
11671         location_t to dump_user_location_t.
11672         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
11673         dump_printf_loc to pass in a dump_location_t rather than a
11674         location_t, via the gimple stmt.
11675         * coverage.c (get_coverage_counts): Update calls to
11676         dump_printf_loc to pass in dump_location_t rather than a
11677         location_t.
11678         * doc/optinfo.texi (Dump types): Convert example of
11679         dump_printf_loc from taking "locus" to taking "insn".  Update
11680         description of the "_loc" calls to cover dump_location_t.
11681         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
11682         "selftest.h".
11683         (dump_user_location_t::dump_user_location_t): New constructors,
11684         from gimple *stmt and rtx_insn *.
11685         (dump_user_location_t::from_function_decl): New function.
11686         (dump_loc): Make static.
11687         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
11688         const dump_location_t &.
11689         (dump_generic_expr_loc): Delete.
11690         (dump_printf_loc): Convert param "loc" from location_t to
11691         const dump_location_t &.
11692         (selftest::test_impl_location): New function.
11693         (selftest::dumpfile_c_tests): New function.
11694         * dumpfile.h: Include "profile-count.h".
11695         (class dump_user_location_t): New class.
11696         (struct dump_impl_location_t): New struct.
11697         (class dump_location_t): New class.
11698         (dump_printf_loc): Convert 2nd param from source_location to
11699         const dump_location_t &.
11700         (dump_generic_expr_loc): Delete.
11701         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
11702         const dump_location_t &.
11703         * gimple-fold.c (fold_gimple_assign): Update call to
11704         dump_printf_loc to pass in a dump_location_t rather than a
11705         location_t, via the gimple stmt.
11706         (gimple_fold_call): Likewise.
11707         * gimple-loop-interchange.cc
11708         (loop_cand::analyze_iloop_reduction_var): Update for change to
11709         check_reduction_path.
11710         (tree_loop_interchange::interchange): Update for change to
11711         find_loop_location.
11712         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
11713         change in return-type of find_loop_location.
11714         (graphite_regenerate_ast_isl): Likewise.
11715         * graphite-optimize-isl.c (optimize_isl): Likewise.
11716         * graphite.c (graphite_transform_loops): Likewise.
11717         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
11718         pass in a dump_location_t rather than a location_t, via the
11719         gimple stmt.
11720         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11721         * ipa.c (walk_polymorphic_call_targets): Likewise.
11722         * loop-unroll.c (report_unroll): Convert "locus" param from
11723         location_t to dump_location_t.
11724         (decide_unrolling): Update for change to get_loop_location's
11725         return type.
11726         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
11727         location_t to dump_user_location_t.
11728         (grid_find_single_omp_among_assignments_1): Updates calls to
11729         dump_printf_loc to pass in a dump_location_t rather than a
11730         location_t, via the gimple stmt.
11731         (grid_parallel_clauses_gridifiable): Convert "tloc" from
11732         location_t to dump_location_t.  Updates calls to dump_printf_loc
11733         to pass in a dump_location_t rather than a location_t, via the
11734         gimple stmt.
11735         (grid_inner_loop_gridifiable_p): Likewise.
11736         (grid_dist_follows_simple_pattern): Likewise.
11737         (grid_gfor_follows_tiling_pattern): Likewise.
11738         (grid_target_follows_gridifiable_pattern): Likewise.
11739         (grid_attempt_target_gridification): Convert initialization
11740         of local "grid" from memset to zero-initialization; FIXME: does
11741         this require C++11?  Update call to dump_printf_loc to pass in a
11742         optinfo_location rather than a location_t, via the gimple stmt.
11743         * profile.c (read_profile_edge_counts): Updates call to
11744         dump_printf_loc to pass in a dump_location_t rather than a
11745         location_t
11746         (compute_branch_probabilities): Likewise.
11747         * selftest-run-tests.c (selftest::run_tests): Call
11748         dumpfile_c_tests.
11749         * selftest.h (dumpfile_c_tests): New decl.
11750         * tree-loop-distribution.c (pass_loop_distribution::execute):
11751         Update for change in return type of find_loop_location.
11752         * tree-parloops.c (parallelize_loops): Likewise.
11753         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
11754         "locus" from location_t to dump_user_location_t.
11755         (canonicalize_loop_induction_variables): Likewise.
11756         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
11757         for change in return type of find_loop_location.
11758         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
11759         to dump_printf_loc to pass in a dump_location_t rather than a
11760         location_t, via the stmt.
11761         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
11762         Likewise.
11763         * tree-vect-loop-manip.c (find_loop_location): Convert return
11764         type from source_location to dump_user_location_t.
11765         (vect_do_peeling): Update for above change.
11766         (vect_loop_versioning): Update for change in type of
11767         vect_location.
11768         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
11769         from location_t to dump_user_location_t.
11770         (vect_estimate_min_profitable_iters): Update for change in type
11771         of vect_location.
11772         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
11773         location_t to dump_location_t.
11774         (vect_slp_bb): Update for change in type of vect_location.
11775         * tree-vectorizer.c (vect_location): Convert from source_location
11776         to dump_user_location_t.
11777         (try_vectorize_loop_1): Update for change in vect_location's type.
11778         (vectorize_loops): Likewise.
11779         (increase_alignment): Likewise.
11780         * tree-vectorizer.h (vect_location): Convert from source_location
11781         to dump_user_location_t.
11782         (find_loop_location): Convert return type from source_location to
11783         dump_user_location_t.
11784         (check_reduction_path): Convert 1st param from location_t to
11785         dump_user_location_t.
11786         * value-prof.c (check_counter): Update call to dump_printf_loc to
11787         pass in a dump_user_location_t rather than a location_t; update
11788         call to error_at for change in type of "locus".
11789         (check_ic_target): Update call to dump_printf_loc to
11790         pass in a dump_user_location_t rather than a location_t, via the
11791         call_stmt.
11793 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11795         * config/s390/s390.h (enum processor_flags): Do not use
11796         default tune parameter when -march was specified.
11798 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
11800         PR target/86314
11801         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
11802         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
11803         operands.
11805 2018-06-26  Richard Biener  <rguenther@suse.de>
11807         PR tree-optimization/86287
11808         PR bootstrap/86316
11809         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
11810         (vect_analyze_loop): Initialize n_stmts.
11812 2018-06-26  Richard Biener  <rguenther@suse.de>
11814         PR middle-end/86271
11815         * fold-const.c (fold_convertible_p): Pointer extension
11816         isn't valid.
11818 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
11820         PR debug/86064
11821         * dwarf2out.c (loc_list_has_views): Adjust comments.
11822         (dw_loc_list): Split single cross-partition range with
11823         nonzero locview.
11825 2018-06-25  Jeff Law  <law@redhat.com>
11827         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
11828         on -mbig-switch by default.
11830         * config/v850/predicates.md (const_float_1_operand): Fix match_code
11831         test.
11832         (const_float_0_operand): Remove unused predicate.
11833         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
11834         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
11835         (recipsf2): New expander.  Original pattern now called
11836         (recipsf2_insn).
11837         (recipdf2, recipdf2_insn): Similarly.
11838         (rsqrtsf2, rsqrtsf2_insn): Similarly
11839         (rsqrtdf2, rsqrtdf2_insn): Similarly
11841 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
11843         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
11844         Simplify logic for FreeBSD (twice).
11846 2018-06-25  Martin Sebor  <msebor@redhat.com>
11848         PR tree-optimization/86204
11849         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
11850         a strnlen result if it's less than the length of the string.
11852 2018-06-25  Martin Sebor  <msebor@redhat.com>
11854         PR tree-optimization/85700
11855         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
11856         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
11857         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
11859 2018-06-25  Martin Sebor  <msebor@redhat.com>
11861         * doc/extend.texi (Zero-length arrays): Update and clarify.
11863 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
11865         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
11866         added IEEE/IBM long double multilib support on PowerPC little
11867         endian Linux systems.
11868         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
11869         (MULTILIB_DEFAULTS): Likewise.
11870         * config/rs6000/rs6000.c (rs6000_option_override_internal):
11871         Likewise.
11872         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
11873         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
11874         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
11876 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
11878         PR middle-end/86311
11879         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
11880         (REORDER_45): Likewise.
11882 2018-06-25  Jeff Law  <law@redhat.com>
11884         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
11885         dividend to 32 bits.  Adjust length.
11886         (udivmodhi4): Cleanup output template.  Fix length.
11888 2018-06-25  Carl Love  <cel@us.ibm.com>
11890         * config/rs6000/vsx.md: Change word selector to prefered location.
11892 2018-06-25  Richard Biener  <rguenther@suse.de>
11894         PR tree-optimization/86304
11895         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
11896         epilogue-if-converted loops as well.
11898 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
11900         * lto-section-out.c (lto_begin_section): Do not print section
11901         name for noaddr and unnumbered dumps.
11903 2018-06-25  Richard Biener  <rguenther@suse.de>
11905         * tree-vectorizer.h (struct vec_info_shared): New structure
11906         with parts split out from struct vec_info and loop_nest from
11907         struct _loop_vec_info.
11908         (struct vec_info): Adjust accordingly.
11909         (struct _loop_vec_info): Likewise.
11910         (LOOP_VINFO_LOOP_NEST): Adjust.
11911         (LOOP_VINFO_DATAREFS): Likewise.
11912         (LOOP_VINFO_DDRS): Likewise.
11913         (struct _bb_vec_info): Likewise.
11914         (BB_VINFO_DATAREFS): Likewise.
11915         (BB_VINFO_DDRS): Likewise.
11916         (struct _stmt_vec_info): Add dr_aux member.
11917         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
11918         (DR_MISALIGNMENT_UNINITIALIZED): New.
11919         (set_dr_misalignment): Adjust.
11920         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
11921         (vect_analyze_loop): Adjust prototype.
11922         (vect_analyze_loop_form): Likewise.
11923         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
11924         Compute dependences lazily.
11925         (vect_record_base_alignments): Use shared datarefs/ddrs.
11926         (vect_verify_datarefs_alignment): Likewise.
11927         (vect_analyze_data_refs_alignment): Likewise.
11928         (vect_analyze_data_ref_accesses): Likewise.
11929         (vect_analyze_data_refs): Likewise.
11930         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
11931         constructor parameter for shared part.
11932         (vect_analyze_loop_form): Pass in shared part and adjust.
11933         (vect_analyze_loop_2): Pass in storage for the number of
11934         stmts.  Move loop nest finding to the caller.  Compute
11935         datarefs lazily.
11936         (vect_analyze_loop): Pass in shared part.
11937         (vect_transform_loop): Verify shared datarefs are unchanged.
11938         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
11939         constructor parameter for shared part.
11940         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
11941         (vect_slp_bb): Verify shared datarefs are unchanged before
11942         transform.
11943         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
11944         change.
11945         (new_stmt_vec_info): Initialize DR_AUX misalignment to
11946         DR_MISALIGNMENT_UNINITIALIZED.
11947         * tree-vectorizer.c (vec_info::vec_info): Add constructor
11948         parameter for shared part.
11949         (vec_info::~vec_info): Adjust.
11950         (vec_info_shared::vec_info_shared): New.
11951         (vec_info_shared::~vec_info_shared): Likewise.
11952         (vec_info_shared::save_datarefs): Likewise.
11953         (vec_info_shared::check_datarefs): Likewise.
11954         (try_vectorize_loop_1): Construct shared part live for analyses
11955         of a single loop for multiple vector sizes.
11956         * tree-parloops.c (gather_scalar_reductions): Adjust.
11958 2018-06-25  Richard Biener  <rguenther@suse.de>
11960         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
11961         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
11962         (vect_analyze_data_refs): Remove similar code from here and
11963         simplify accordingly.
11965 2018-06-25  Richard Biener  <rguenther@suse.de>
11967         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
11968         for reverse storage order accesses rather than asserting
11969         they cannot happen here.
11971 2018-06-25  Tom de Vries  <tdevries@suse.de>
11973         PR debug/86257
11974         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
11975         Use data16 instead of .byte for insn prefix.
11977 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
11979         PR C++/86082
11980         * parser.c (make_char_string_pack): Pass this literal chars
11981         through cpp_interpret_string.
11982         (cp_parser_userdef_numeric_literal): Check the result of
11983         make_char_string_pack.
11985 2018-06-24  Maya Rashish  <coypu@sdf.org>
11987         * ginclude/stddef.h: Simplify conditions around avoiding
11988         re-definition of __size_t.
11990 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
11992         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
11993         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
11995 2018-06-22  Maya Rashish  <coypu@sdf.org>
11997         * doc/invoke.texi (mno-fancy-math-387): Update for changes
11998         made to OpenBSD and NetBSD through the years.
12000 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12002         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12003         behavior of vec_pack (vector double, vector double) to match
12004         behavior of vec_float2 (vector double, vector double).
12006 2018-06-22  Olivier Hainque  <hainque@adacore.com>
12008         * gimplify.c (gimplify_function_tree): Prevent creation
12009         of a trampoline for the address of the current function
12010         passed to entry/exit instrumentation hooks.
12012 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
12014         PR target/86222
12015         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
12016         correctly.
12018 2018-06-22  Martin Liska  <mliska@suse.cz>
12020         PR tree-optimization/86263
12021         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12022         Bail out if is_enabled is false.
12023         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
12024         New declaration.
12025         (jump_table_cluster::is_enabled): New function.
12027 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12029         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
12030         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
12031         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
12032         (lto_input_ts_binfo_tree_pointers): Likewise.
12033         * tree-streamer-out.c (streamer_write_tree_bitfields,
12034         write_ts_binfo_tree_pointers): Likewise.
12035         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
12037 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
12039         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
12041 2018-06-22  Martin Liska  <mliska@suse.cz>
12043         * symbol-summary.h (get): Make it pure and inline move
12044         functionality from ::get function.
12045         (get): Remove and inline into ::get and ::get_create.
12046         (get_create): Move code from ::get function.
12048 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12050         PR target/85994
12051         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
12052         -x assembler-with-cpp.
12054 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12056         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
12057         _FILE_OFFSET_BITS=64 for C++.
12059 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
12061         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
12062         conversion insn that shows up when pr85657-3.c is compiled using
12063         IEEE 128-bit long double.
12064         (neg<mode>2_internal): Use the correct mode to check whether the
12065         mode is IBM extended.
12066         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
12067         multiply and divide external functions from being created more
12068         than once.
12070 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12072         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
12073         functions.
12074         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
12075         the edge can be forwarded.
12076         (cfg_layout_merge_blocks): Likewise.
12078 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12080         * except.c (finish_eh_generation): Commit edge insertions only after
12081         the EH edges have been redirected from post-landing to landing pads.
12083 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12085         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
12086         create_tmp_var_for to create the FRAME decl.
12087         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
12089 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
12091         * tree-inline.c (copy_edges_for_bb): Minor tweak.
12092         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
12093         debug statement when resetting its value.
12094         (expand_call_inline): Copy the locus of the call onto the assignment
12095         of the return value, if any.  Use local variable in more cases.
12097 2018-06-21  Martin Liska  <mliska@suse.cz>
12099         * ipa-pure-const.c (propagate_nothrow): Use
12100         funct_state_summaries->get.
12101         (dump_malloc_lattice): Likewise.
12102         (propagate_malloc): Likewise.
12104 2018-06-21  Richard Biener  <rguenther@suse.de>
12106         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
12107         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
12108         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
12109         comment.
12110         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
12111         BLOCK_ABSTRACT_ORIGIN unconditionally.
12113 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
12115         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
12116         deleting it.
12117         * ipa-reference.c (ipa_reference_c_finalize): Delete
12118         ipa_ref_opt_sum_summaries and set it to NULL.
12120 2018-06-21  Tom de Vries  <tdevries@suse.de>
12122         PR tree-optimization/85859
12123         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
12124         test with comment from bb_no_side_effects_p.
12126 2018-06-21  Richard Biener  <rguenther@suse.de>
12128         PR tree-optimization/86232
12129         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
12130         max for constant niter.
12132 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12134         * config/aarch64/aarch64-simd.md
12135         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
12137 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12139         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
12140         Make opernads of the unspec commutative.
12142 2018-06-21  Richard Biener  <rguenther@suse.de>
12144         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
12145         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12146         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
12147         (vect_analyze_data_ref_dependence): Re-order checks to deal with
12148         NULL DR_STEP.
12149         (vect_record_base_alignments): Do not record base alignment
12150         for gathers or scatters.
12151         (vect_compute_data_ref_alignment): Drop return value that is always
12152         true.  Bail out early for gathers or scatters.
12153         (vect_enhance_data_refs_alignment): Bail out early for gathers
12154         or scatters.
12155         (vect_find_same_alignment_drs): Likewise.
12156         (vect_analyze_data_refs_alignment): Remove dead code.
12157         (vect_slp_analyze_and_verify_node_alignment): Likewise.
12158         (vect_analyze_data_refs): For possible gathers or scatters do
12159         not create an alternate DR, just check their possible validity
12160         and mark them.  Adjust DECL_NONALIASED handling to not rely
12161         on DR_BASE_ADDRESS.
12162         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
12163         update inits of gathers or scatters.
12164         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
12165         Also copy gather/scatter flag to pattern vinfo.
12167 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12169         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
12170         behavior of vec_packsu (vector unsigned long long, vector unsigned
12171         long long) to match behavior of vec_packs with same signature.
12173 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
12174             Thomas Schwinge <thomas@codesourcery.com>
12175             Cesar Philippidis  <cesar@codesourcery.com>
12177         * gimplify.c (gimplify_scan_omp_clauses): Add support for
12178         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
12179         (gimplify_adjust_omp_clauses): Likewise.
12180         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
12181         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
12182         (gimplify_omp_target_update): Update handling of acc update and
12183         enter/exit data.
12184         * omp-low.c (install_var_field): Remove unused parameter
12185         base_pointers_restrict.
12186         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
12187         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
12188         FINALIZE}
12189         (omp_target_base_pointers_restrict_p): Delete.
12190         (scan_omp_target): Update call to scan_sharing_clauses.
12191         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
12192         FINALIZE}.
12193         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
12194         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
12195         (convert_local_omp_clauses): Likewise.
12196         * tree-pretty-print.c (dump_omp_clause): Likewise.
12197         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
12198         FINALIZE}.
12199         (omp_clause_code_name): Likewise.
12201 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
12203         PR debug/86194
12204         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
12205         be narrowed.
12207         PR tree-optimization/86231
12208         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
12209         anti-range don't overwrite *vr0min before using it to compute *vr0max.
12211 2018-06-20  Tom de Vries  <tdevries@suse.de>
12213         PR tree-optimization/86097
12214         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
12215         iv type if signedness of iv type is not the same as that of *nit.
12217 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
12219         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
12220         EDGE_EH edges, verify they are all EDGE_EH.
12222 2018-06-20  Maya Rashish  <coypu@sdf.org>
12224         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
12226 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12228         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
12229         * config/aarch64/aarch64.c (xgene1_tunings): Add
12230         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
12231         (aarch64_mode_valid_for_sched_fusion_p):
12232         Allow 16-byte modes.
12233         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
12234         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
12235         128-bit modes.
12236         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
12237         New pattern.
12238         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
12239         * config/aarch64/iterators.md (VQ2): New mode iterator.
12241 2018-06-20  Martin Liska  <mliska@suse.cz>
12243         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
12244         Change default ratio from 10 to 8.
12246 2018-06-20  Martin Liska  <mliska@suse.cz>
12248         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
12249         New.
12250         (bit_test_cluster::find_bit_tests): Likewise.
12251         (switch_decision_tree::analyze_switch_statement): Find clusters.
12252         * tree-switch-conversion.h (struct jump_table_cluster): Document
12253         hierarchy.
12255 2018-06-20  Martin Liska  <mliska@suse.cz>
12257         * tree-switch-conversion.c (switch_conversion::collect):
12258         Record m_uniq property.
12259         (switch_conversion::expand): Bail out for special conditions.
12260         (group_cluster::~group_cluster): New.
12261         (group_cluster::group_cluster): Likewise.
12262         (group_cluster::dump): Likewise.
12263         (jump_table_cluster::emit): New.
12264         (switch_decision_tree::fix_phi_operands_for_edges): New.
12265         (struct case_node): Remove struct.
12266         (jump_table_cluster::can_be_handled): New.
12267         (case_values_threshold): Moved to header.
12268         (reset_out_edges_aux): Likewise.
12269         (jump_table_cluster::is_beneficial): New.
12270         (bit_test_cluster::can_be_handled): Likewise.
12271         (add_case_node): Remove.
12272         (bit_test_cluster::is_beneficial): New.
12273         (case_bit_test::cmp): New.
12274         (bit_test_cluster::emit): New.
12275         (expand_switch_as_decision_tree_p): Remove.
12276         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
12277         (fix_phi_operands_for_edge): Likewise.
12278         (switch_decision_tree::analyze_switch_statement): New.
12279         (compute_cases_per_edge): Move ...
12280         (switch_decision_tree::compute_cases_per_edge): ... here.
12281         (try_switch_expansion): Likewise.
12282         (switch_decision_tree::try_switch_expansion): Likewise.
12283         (record_phi_operand_mapping): Likewise.
12284         (switch_decision_tree::record_phi_operand_mapping): Likewise.
12285         (emit_case_decision_tree): Likewise.
12286         (switch_decision_tree::emit): Likewise.
12287         (balance_case_nodes): Likewise.
12288         (switch_decision_tree::balance_case_nodes): Likewise.
12289         (dump_case_nodes): Likewise.
12290         (switch_decision_tree::dump_case_nodes): Likewise.
12291         (emit_jump): Likewise.
12292         (switch_decision_tree::emit_jump): Likewise.
12293         (emit_cmp_and_jump_insns): Likewise.
12294         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
12295         (emit_case_nodes): Likewise.
12296         (switch_decision_tree::emit_case_nodes): Likewise.
12297         (conditional_probability): Remove.
12298         * tree-switch-conversion.h (enum cluster_type): New.
12299         (PRINT_CASE): New.
12300         (struct cluster): Likewise.
12301         (cluster::cluster): Likewise.
12302         (struct simple_cluster): Likewise.
12303         (simple_cluster::simple_cluster): Likewise.
12304         (struct group_cluster): Likewise.
12305         (struct jump_table_cluster): Likewise.
12306         (struct bit_test_cluster): Likewise.
12307         (struct min_cluster_item): Likewise.
12308         (struct case_tree_node): Likewise.
12309         (case_tree_node::case_tree_node): Likewise.
12310         (jump_table_cluster::case_values_threshold): Likewise.
12311         (struct case_bit_test): Likewise.
12312         (struct switch_decision_tree): Likewise.
12313         (struct switch_conversion): Likewise.
12314         (switch_decision_tree::reset_out_edges_aux): Likewise.
12316 2018-06-20  Martin Liska  <mliska@suse.cz>
12318         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
12319         (hoist_edge_and_branch_if_true): Likewise.
12320         (expand_switch_using_bit_tests_p): Likewise.
12321         (struct case_bit_test): Likewise.
12322         (case_bit_test_cmp): Likewise.
12323         (emit_case_bit_tests): Likewise.
12324         (switch_conversion::switch_conversion): New class.
12325         (struct switch_conv_info): Remove old struct.
12326         (collect_switch_conv_info): More to ...
12327         (switch_conversion::collect): ... this.
12328         (check_range): Likewise.
12329         (switch_conversion::check_range): Likewise.
12330         (check_all_empty_except_final): Likewise.
12331         (switch_conversion::check_all_empty_except_final): Likewise.
12332         (check_final_bb): Likewise.
12333         (switch_conversion::check_final_bb): Likewise.
12334         (create_temp_arrays): Likewise.
12335         (switch_conversion::create_temp_arrays): Likewise.
12336         (free_temp_arrays): Likewise.
12337         (gather_default_values): Likewise.
12338         (switch_conversion::gather_default_values): Likewise.
12339         (build_constructors): Likewise.
12340         (switch_conversion::build_constructors): Likewise.
12341         (constructor_contains_same_values_p): Likewise.
12342         (switch_conversion::contains_same_values_p): Likewise.
12343         (array_value_type): Likewise.
12344         (switch_conversion::array_value_type): Likewise.
12345         (build_one_array): Likewise.
12346         (switch_conversion::build_one_array): Likewise.
12347         (build_arrays): Likewise.
12348         (switch_conversion::build_arrays): Likewise.
12349         (gen_def_assigns): Likewise.
12350         (switch_conversion::gen_def_assigns): Likewise.
12351         (prune_bbs): Likewise.
12352         (switch_conversion::prune_bbs): Likewise.
12353         (fix_phi_nodes): Likewise.
12354         (switch_conversion::fix_phi_nodes): Likewise.
12355         (gen_inbound_check): Likewise.
12356         (switch_conversion::gen_inbound_check): Likewise.
12357         (process_switch): Use the newly created class.
12358         (switch_conversion::expand): New.
12359         (switch_conversion::~switch_conversion): New.
12360         * tree-switch-conversion.h: New file.
12362 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12364         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
12365         tree-vect-patterns.c.
12366         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
12367         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
12368         (vect_recog_sad_pattern): Likewise.
12369         (vect_recog_widen_sum_pattern): Likewise.
12370         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
12371         (vect_recog_widen_shift_pattern): Remove the type_in argument.
12372         (vect_recog_rotate_pattern): Likewise.
12373         (vect_recog_mult_pattern): Likewise.
12374         (vect_recog_vector_vector_shift_pattern): Likewise.
12375         (vect_recog_divmod_pattern): Likewise.
12376         (vect_recog_mixed_size_cond_pattern): Likewise.
12377         (vect_recog_bool_pattern): Likewise.
12378         (vect_recog_mask_conversion_pattern): Likewise.
12379         (vect_try_gather_scatter_pattern): Likewise.
12380         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
12381         (vect_recog_over_widening_pattern): Likewise.
12382         (vect_recog_gather_scatter_pattern): Likewise.
12383         (vect_recog_func_ptr): Move from tree-vectorizer.h
12384         (vect_vect_recog_func_ptrs): Move further down the file.
12385         (vect_recog_func): Likewise.  Remove the third argument.
12386         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
12387         (vect_pattern_recog_1): Expect the pattern function to do any
12388         necessary target tests.  Also expect it to provide a vector type.
12389         Remove the type_in handling.
12391 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12393         * tree-vect-patterns.c (vect_pattern_detected): New function.
12394         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
12395         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
12396         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
12397         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
12398         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
12399         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
12400         (vect_recog_mask_conversion_pattern)
12401         (vect_try_gather_scatter_pattern): Likewise.
12403 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12405         * tree-vect-patterns.c (vect_get_internal_def): New function.
12406         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
12407         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
12408         (search_type_for_mask_1): Use it.
12410 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12412         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
12413         redundant WIDEN_SUM_EXPR handling.
12414         (vect_recog_sad_pattern): Likewise.
12416 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12418         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
12419         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
12420         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
12421         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
12422         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
12424 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12426         * tree-vect-stmts.c (vectorizable_call): Make sure that we
12427         use the stmt_vec_info of the original bb statement for the
12428         new zero assignment, even if the call is part of a pattern.
12430 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12432         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
12433         that the sequence is attached to the original statement rather
12434         than the pattern statement.
12435         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
12436         PATTERN_DEF_SEQ from the original statement rather than
12437         the main pattern statement.
12438         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
12439         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
12440         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
12442 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
12444         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
12445         definition statements before the early exit for statements that aren't
12446         live or relevant.
12447         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
12448         split out from...
12449         (vect_transform_loop): ...here.  Process pattern definition
12450         statements without first checking whether the main pattern
12451         statement is live or relevant.
12453 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
12455         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
12456         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
12458 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
12460         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
12461         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
12462         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
12463         (expand_block_compare): Change select_block_compare_mode call.
12464         (expand_strncmp_align_check): Use new functions, fix comment.
12465         (emit_final_str_compare_gpr): New function.
12466         (expand_strn_compare): Refactor and clean up code.
12467         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
12469 2018-06-19  Tony Reix  <tony.reix@atos.com>
12470             Damien Bergamini  <damien.bergamini@atos.com>
12471             David Edelsohn  <dje.gcc@gmail.com>
12473         * collect2.c (static_obj): New variable.
12474         (static_libs): New variable.
12475         (is_in_list): Uncomment declaration.
12476         (main): Track AIX libraries linked statically.
12477         (is_in_list): Uncomment definition.
12478         (scan_prog_file): Don't add AIX shared libraries initializer
12479         to constructor list if linking statically.
12481 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
12483         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
12484         constant.
12485         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
12487 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
12489         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
12490         blocks.
12492 2018-06-19  Martin Liska  <mliska@suse.cz>
12494         * config/i386/i386.c (ix86_can_inline_p): Do not use
12495         ipa_fn_summaries::get_create.
12496         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
12497         get.
12498         (devirtualization_time_bonus): Likewise.
12499         (ipcp_propagate_stage): Likewise.
12500         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
12501         (edge_set_predicate): Likewise.
12502         (evaluate_conditions_for_known_args): Likewise.
12503         (evaluate_properties_for_edge): Likewise.
12504         (ipa_call_summary::reset): Tranform to ...
12505         (ipa_call_summary::~ipa_call_summary): ... this.
12506         (ipa_fn_summary::reset): Transform to ...
12507         (ipa_fn_summary::~ipa_fn_summary): ... this.
12508         (ipa_fn_summary_t::remove): Rename to ...
12509         (ipa_fn_summary_t::remove_callees): ... this.
12510         (ipa_fn_summary_t::duplicate): Use placement new
12511         instead of memory copy.
12512         (ipa_call_summary_t::duplicate): Likewise.
12513         (ipa_call_summary_t::remove): Remove.
12514         (dump_ipa_call_summary): Change get_create to get.
12515         (ipa_dump_fn_summary): Dump only when summary exists.
12516         (analyze_function_body): Use symbol_summary::get instead
12517         of get_create.
12518         (compute_fn_summary): Likewise.
12519         (estimate_edge_devirt_benefit): Likewise.
12520         (estimate_edge_size_and_time): Likewise.
12521         (inline_update_callee_summaries): Likewise.
12522         (remap_edge_change_prob): Likewise.
12523         (remap_edge_summaries): Likewise.
12524         (ipa_merge_fn_summary_after_inlining): Likewise.
12525         (write_ipa_call_summary): Likewise.
12526         (ipa_fn_summary_write): Likewise.
12527         (ipa_free_fn_summary): Likewise.
12528         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
12529         (struct ipa_call_summary): Likewise.
12530         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
12531         of get_create.
12532         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
12533         (estimate_size_after_inlining): Likewise.
12534         (estimate_growth): Likewise.
12535         (growth_likely_positive): Likewise.
12536         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
12537         (inline_call): Likewise.
12538         * ipa-inline.c (caller_growth_limits): Likewise.
12539         (can_inline_edge_p): Likewise.
12540         (can_inline_edge_by_limits_p): Likewise.
12541         (compute_uninlined_call_time): Likewise.
12542         (compute_inlined_call_time): Likewise.
12543         (want_inline_small_function_p): Likewise.
12544         (edge_badness): Likewise.
12545         (update_caller_keys): Likewise.
12546         (update_callee_keys): Likewise.
12547         (inline_small_functions): Likewise.
12548         (inline_to_all_callers_1): Likewise.
12549         (dump_overall_stats): Likewise.
12550         (early_inline_small_functions): Likewise.
12551         (early_inliner): Likewise.
12552         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
12553         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12554         * ipa-pure-const.c (malloc_candidate_p): Likewise.
12555         * ipa-split.c (execute_split_functions): Likewise.
12556         * symbol-summary.h: Likewise.
12557         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
12559 2018-06-19  Richard Biener  <rguenther@suse.de>
12561         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
12562         (vectorize_loops): ... here.  Fix dbgcnt handling.
12563         (try_vectorize_loop): Wrap try_vectorize_loop_1.
12565 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
12567         PR target/86197
12568         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
12569         ieee128 argument takes up only one (vector) register, not two (floating
12570         point) registers.
12572 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
12574         * gimplify.c (gimplify_init_constructor): Really never clear for an
12575         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
12577 2018-06-19  Richard Biener  <rguenther@suse.de>
12579         PR tree-optimization/86179
12580         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
12581         after failed recognition.
12583 2018-06-18  Martin Sebor  <msebor@redhat.com>
12585         PR middle-end/85602
12586         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
12587         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
12588         Handle integer subtraction.
12589         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
12590         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
12592 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
12594         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
12595         param from rtx to rtx_insn *.
12596         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
12597         param.
12598         (frv_ifcvt_modify_insn): Likwise.
12599         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
12600         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
12601         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
12602         as_a <rtx_insn *> cast to local "unprotected_region" once
12603         it's been established that it's not NULL or pc_rtx.
12604         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
12605         param "sethi" from rtx to rtx_insn *.
12606         (nds32_group_float_insns): Likewise for param "insn".
12607         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
12608         param.
12609         (vax_output_int_subtract): Likewise.
12610         * config/vax/vax.c (vax_output_int_add): Likewise for param
12611         "insn".
12612         (vax_output_int_subtract): Likewise.
12613         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
12614         (emit_pattern_after): Likewise for param "after".
12615         (emit_insn_after): Likewise.
12616         (emit_jump_insn_after): Likewise.
12617         (emit_call_insn_after): Likewise.
12618         (emit_debug_insn_after): Likewise.
12619         (emit_pattern_before): Likewise for param "before".
12620         (emit_insn_before): Likewise.
12621         (emit_jump_insn_before): Likewise.
12622         * final.c (get_insn_template): Likewise for param "insn", removing
12623         a cast.
12624         * output.h (get_insn_template): Likewise for 2nd param.
12625         * rtl.h (emit_insn_before): Likewise.
12626         (emit_jump_insn_before): Likewise.
12627         (emit_debug_insn_before_noloc): Likewise.
12628         (emit_insn_after): Likewise.
12629         (emit_jump_insn_after): Likewise.
12630         (emit_call_insn_after): Likewise.
12631         (emit_debug_insn_after): Likewise.
12632         (set_insn_deleted): Likewise for param.
12634 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
12636         PR target/85358
12637         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
12638         floating point modes, so that IFmode is numerically greater than
12639         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
12640         to declare the ordering.  This prevents IFmode from being
12641         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
12642         machine.  Include rs6000-modes.h to share the fractional values
12643         between genmodes* and the rest of the compiler.
12644         (IFmode): Likewise.
12645         (KFmode): Likewise.
12646         (TFmode): Likewise.
12647         * config/rs6000/rs6000-modes.h: New file.
12648         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
12649         meaning of rs6000_long_double_size so that 126..128 selects an
12650         appropriate 128-bit floating point type.
12651         (rs6000_option_override_internal): Likewise.
12652         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
12653         (TARGET_LONG_DOUBLE_128): Change the meaning of
12654         rs6000_long_double_size so that 126..128 selects an appropriate
12655         128-bit floating point type.
12656         (LONG_DOUBLE_TYPE_SIZE): Update comment.
12657         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
12658         source and destination to match the standard usage.
12659         (truncifkf2): Likewise.
12660         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
12661         ISA 2.07 to use an explicit clobber, instead of passing in a
12662         temporary.
12663         (copysign<mode>3_soft): Likewise.
12665 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
12667         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
12668         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
12669         (vect_slp_analyze_instance_dependence): Likewise.
12670         (vect_enhance_data_refs_alignment): Likewise.
12671         (vect_analyze_data_refs_alignment): Likewise.
12672         (vect_slp_analyze_and_verify_instance_alignment
12673         (vect_analyze_data_ref_accesses): Likewise.
12674         (vect_prune_runtime_alias_test_list): Likewise.
12675         (vect_analyze_data_refs): Likewise.
12676         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
12677         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12678         (vect_analyze_scalar_cycles_1): Likewise.
12679         (vect_get_loop_niters): Likewise.
12680         (vect_analyze_loop_form_1): Likewise.
12681         (vect_update_vf_for_slp): Likewise.
12682         (vect_analyze_loop_operations): Likewise.
12683         (vect_analyze_loop): Likewise.
12684         (vectorizable_induction): Likewise.
12685         (vect_transform_loop): Likewise.
12686         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
12687         * tree-vect-slp.c (vect_analyze_slp): Likewise.
12688         (vect_make_slp_decision): Likewise.
12689         (vect_detect_hybrid_slp): Likewise.
12690         (vect_slp_analyze_operations): Likewise.
12691         (vect_slp_bb): Likewise.
12692         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
12693         (vectorizable_bswap): Likewise.
12694         (vectorizable_call): Likewise.
12695         (vectorizable_simd_clone_call): Likewise.
12696         (vectorizable_conversion): Likewise.
12697         (vectorizable_assignment): Likewise.
12698         (vectorizable_shift): Likewise.
12699         (vectorizable_operation): Likewise.
12700         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
12702 2018-06-18  Martin Sebor  <msebor@redhat.com>
12704         PR tree-optimization/81384
12705         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
12706         * builtins.c (expand_builtin_strnlen): New function.
12707         (expand_builtin): Call it.
12708         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
12709         * builtins.def (BUILT_IN_STRNLEN): New.
12710         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
12711         Warn for bounds in excess of maximum object size.
12712         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
12713         single-value ranges.  Handle strnlen.
12714         (handle_builtin_strlen): Handle strnlen.
12715         (strlen_check_and_optimize_stmt): Same.
12716         * doc/extend.texi (Other Builtins): Document strnlen.
12718 2018-06-18  Maya Rashish  <coypu@sdf.org>
12720         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
12721         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
12722         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
12724         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
12725         here to ...
12726         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
12728 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12730         * tree.c (escaped_string::escape): Replace cast to char * by
12731         const_cast<char *> (unescaped).
12733 2018-06-18  Nick Clifton  <nickc@redhat.com>
12735         PR 84195
12736         * tree.c (escaped_string): New class.  Converts an unescaped
12737         string into its escaped equivalent.
12738         (warn_deprecated_use): Use the new class to convert the
12739         deprecation message, if present.
12740         (test_escaped_strings): New self test.
12741         (test_c_tests): Add test_escaped_strings.
12742         * doc/extend.texi (deprecated): Add a note that the
12743         deprecation message is affected by the -fmessage-length
12744         option, and that control characters will be escaped.
12745         (#pragma GCC error): Document this pragma.
12746         (#pragma GCC warning): Likewise.
12747         * doc/invoke.texi (-fmessage-length): Document this option's
12748         effect on the #warning and #error preprocessor directives and
12749         the deprecated attribute.
12751 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
12753         * tree.c (decl_value_expr_lookup): Revert latest change.
12754         (decl_value_expr_insert): Likewise.
12756 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
12758         * gimplify.c (nonlocal_vlas): Delete.
12759         (nonlocal_vla_vars): Likewise.
12760         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
12761         referenced VLAs.
12762         (gimplify_body): Do not create and destroy nonlocal_vlas.
12763         * tree-nested.c: Include diagnostic.h.
12764         (use_pointer_in_frame): Tweak.
12765         (lookup_field_for_decl): Add assertion and declare the transformation.
12766         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
12767         internal error when the reference is in a wrong context.  Do not
12768         create a debug decl by default.
12769         (note_nonlocal_block_vlas): Delete.
12770         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
12771         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
12772         create a debug decl by default.
12773         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
12774         call is in a wrong context.
12775         (fixup_vla_decls): New function.
12776         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
12777         debug variables were created.
12778         * tree.c (decl_value_expr_lookup): Add checking assertion.
12779         (decl_value_expr_insert): Likewise.
12781 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
12783         PR middle-end/82479
12784         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
12785         * tree-scalar-evolution.c (interpret_expr): Likewise.
12786         (expression_expensive_p): Likewise.
12787         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
12788         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
12789         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
12790         (ssa_defined_by_minus_one_stmt_p): New.
12792 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
12794         PR middle-end/64946
12795         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
12796         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
12797         * dojump.c (do_jump): Likewise.
12798         * expr.c (expand_expr_real_2): Check operand type's sign.
12799         * fold-const.c (const_unop): Handle ABSU_EXPR.
12800         (fold_abs_const): Likewise.
12801         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
12802         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
12803         (strip_sign_op_1): Likesise.
12804         * match.pd: Add new pattern to generate ABSU_EXPR.
12805         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
12806         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
12807         * tree-eh.c (operation_could_trap_helper_p): Likewise.
12808         * tree-inline.c (estimate_operator_cost): Likewise.
12809         * tree-pretty-print.c (dump_generic_node): Likewise.
12810         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
12811         * tree.def (ABSU_EXPR): New.
12813 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
12815         PR middle-end/86095
12816         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
12817         documented as preserved for backward compatibility only.
12818         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
12820         PR rtl-optimization/86108
12821         * bb-reorder.c (create_forwarder_block): Renamed to ...
12822         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
12823         jump from new landing pad to the second part.
12824         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
12825         Adjust callers.
12827 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
12829         PR middle-end/85878
12830         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
12831         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
12832         Only call store_expr for halves if the mode is the same.
12834         PR middle-end/86123
12835         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
12836         Fix up comment formatting.
12838 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12840         * typed-splay-tree.h (typed_splay_tree::remove): New function.
12841         (typed_splay_tree::closure,
12842         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
12843         (typed_splay_tree::typed_splay_tree,
12844         typed_splay_tree::operator =): Declared private.
12845         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
12846         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
12847         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
12848         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
12849         typed_splay_tree::splay_tree_splay,
12850         typed_splay_tree::splay_tree_foreach_helper,
12851         typed_splay_tree::splay_tree_insert,
12852         typed_splay_tree::splay_tree_remove,
12853         typed_splay_tree::splay_tree_lookup,
12854         typed_splay_tree::splay_tree_predecessor,
12855         typed_splay_tree::splay_tree_successor,
12856         typed_splay_tree::splay_tree_min,
12857         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
12858         (typed_splay_tree::root, typed_splay_tree::comp,
12859         typed_splay_tree::delete_key,
12860         typed_splay_tree::delete_value): New data members.
12861         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
12862         typed_splay_tree::remove.
12864 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
12866         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
12867         -mginv and -mno-ginv to the assembler.
12868         * config/mips/mips.opt (-mcrc): New option.
12869         (-mginv): Likewise.
12870         * doc/invoke.text (-mcrc): Document.
12871         (-mginv): Likewise.
12873 2018-06-15  Nick Clifton  <nickc@redhat.com>
12875         PR 84195
12876         * tree.c (escaped_string): New class.  Converts an unescaped
12877         string into its escaped equivalent.
12878         (warn_deprecated_use): Use the new class to convert the
12879         deprecation message, if present.
12880         (test_escaped_strings): New self test.
12881         (test_c_tests): Add test_escaped_strings.
12882         * doc/extend.texi (deprecated): Add a note that the
12883         deprecation message is affected by the -fmessage-length
12884         option, and that control characters will be escaped.
12885         (#pragma GCC error): Document this pragma.
12886         (#pragma GCC warning): Likewise.
12887         * doc/invoke.texi (-fmessage-length): Document this option's
12888         effect on the #warning and #error preprocessor directives and
12889         the deprecated attribute.
12891 2018-06-15  Richard Biener  <rguenther@suse.de>
12893         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
12894         here, also noting vector size used.
12895         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
12896         size used in MSG_OPTIMIZED_LOCATIONS dump.
12897         (pass_slp_vectorize::execute): Adjust.
12899 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
12901         PR target/85968
12902         * config/arc/arc.c (arc_return_address_register): Fix
12903         if-condition.
12905 2018-06-15  Richard Biener  <rguenther@suse.de>
12907         PR middle-end/86159
12908         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
12909         leave useless conversion stripping to force_gimple_operand_gsi.
12910         (gimplify_build2): Likewise.
12911         (gimplify_build1): Likewise.
12913 2018-06-15  Richard Biener  <rguenther@suse.de>
12915         PR middle-end/86076
12916         * tree-cfg.c (move_stmt_op): unshare invariant addresses
12917         before adjusting their block.
12919 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12921         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
12922         multilibs for *-*-rtems*.
12923         * config/riscv/t-rtems: New file.
12925 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
12927         PR middle-end/86122
12928         * match.pd ((A +- CST1) +- CST2): Punt if last resort
12929         unsigned_type_for returns NULL.
12931         PR target/85945
12932         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
12933         subregs of multi-word pseudos unless the float mode has word size.
12935 2018-06-14  Richard Biener  <rguenther@suse.de>
12937         PR middle-end/86139
12938         * tree-vect-generic.c (build_word_mode_vector_type): Remove
12939         duplicate and harmful type_hash_canon.
12940         * tree.c (type_hash_canon): Assert we didn't find ourselves.
12942 2018-06-14  Richard Biener  <rguenther@suse.de>
12944         PR ipa/86124
12945         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
12946         NULL cgraph_node.
12948 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12950         * config/rtems.h (STDINT_LONG32): Define.
12952 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
12953             Prachi Godbole  <prachi.godbole@imgtec.com>
12955         * config/mips/mips-cpus.def: Define P6600.
12956         * config/mips/mips-tables.opt: Regenerate.
12957         * config/mips/mips.c (mips_ucbranch_type): New enum.
12958         (mips_rtx_cost_data): Add support for P6600.
12959         (mips_issue_rate): Likewise.
12960         (mips_multipass_dfa_lookahead): Likewise.
12961         (mips_avoid_hazard): Likewise.
12962         (mips_reorg_process_insns): Likewise.
12963         (mips_classify_branch_p6600): New function.
12964         * config/mips/mips.h (TUNE_P6600): New define.
12965         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
12966         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
12967         * config/mips/mips.md: Include p6600.md.
12968         (processor): Add p6600.
12969         * config/mips/p6600.md: New file.
12970         * doc/invoke.texi: Add p6600 to supported architectures.
12972 2018-06-13  Martin Sebor  <msebor@redhat.com>
12974         PR tree-optimization/86114
12975         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
12976         of integer types.
12977         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
12979 2018-06-13  Richard Biener  <rguenther@suse.de>
12981         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
12982         Properly set vector type of the intermediate stmt.
12983         * tree-vect-stmts.c (vectorizable_operation): The destination
12984         var always has vectype_out type.
12986 2018-06-13  Jeff Law  <law@redhat.com>
12988         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
12989         integer 0 for argument to print_rtl_with_bb.
12990         (rl78_reorg): Likewise.
12992 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
12994         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
12995         from rtx to rtx_insn *.
12996         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
12997         "label".
12998         (add_sched_insns_for_speculation): Likewise for local "target",
12999         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13000         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
13001         from rtx_insn ** to rtx_code_label **.
13002         (reorg_emit_nops): Likewise.
13003         (c6x_reorg): Likewise for local "call_labels".
13004         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
13005         rtx to rtx_insn *.
13006         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
13007         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
13008         the loops over LABEL_REFS.
13009         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
13010         braf_label.
13011         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
13012         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
13013         (split_branches): Strengthen local "olabel" from rtx to
13014         rtx_insn *, adding a safe_as_a cast.
13015         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
13016         to "rtx_insn *".
13017         (add_insn_after): Likewise for first two params.
13018         (add_insn_before): Likewise.
13019         (remove_insn): Likewise for param.
13020         (emit_pattern_before_noloc): Likewise for second and third params.
13021         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
13022         (emit_call_insn_before_noloc): Likewise.
13023         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
13024         to "rtx_insn *".
13025         (emit_barrier_before): Likewise.
13026         (emit_label_before): Strengthen "label" param from "rtx" to
13027         "rtx_code_label *".  Strengthen "before" param from "rtx" to
13028         "rtx_insn *".
13029         (emit_insn_after_1): Strengthen "after" param from "rtx" to
13030         "rtx_insn *".
13031         (emit_pattern_after_noloc): Likewise.
13032         (emit_insn_after_noloc): Likewise.
13033         (emit_jump_insn_after_noloc): Likewise.
13034         (emit_call_insn_after_noloc): Likewise.
13035         (emit_debug_insn_after_noloc): Likewise.
13036         (emit_barrier_after): Likewise.
13037         (emit_label_after): Likewise for both params.
13038         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
13039         "loc" param from "int" to "location_t".
13040         (emit_insn_after_setloc): Likewise.
13041         (emit_jump_insn_after_setloc): Likewise.
13042         (emit_call_insn_after_setloc): Likewise.
13043         (emit_debug_insn_after_setloc): Likewise.
13044         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
13045         "loc" param from "int" to "location_t".
13046         (emit_pattern_before): Convert NULL_RTX to NULL.
13047         (emit_insn_before_setloc): Convert "loc" param from "int" to
13048         "location_t".
13049         (emit_jump_insn_before_setloc): Likewise.
13050         (emit_call_insn_before_setloc): Likewise.
13051         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
13052         rtx_insn *.  Convert "loc" param from "int" to "location_t".
13053         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
13054         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
13055         Convert 3rd param from "int" to "location_t".
13056         (emit_barrier_before, emit_barrier_after, next_real_insn):
13057         Strengthen param from rtx to rtx_insn *.
13058         (emit_label_before): Strengthen 1st param from "rtx" to
13059         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
13060         "rtx_insn *".
13061         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
13062         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
13063         Strengthen 2nd param from "rtx" to "rtx_insn *".
13064         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
13065         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
13066         Likewise. Convert 3rd param from "int" to "location_t".
13067         (emit_label_after): Strengthen 1st param from "rtx" to
13068         "rtx_code_label *".
13069         (next_real_insn, remove_insn): Strengthen param from "rtx" to
13070         "rtx_insn *".
13071         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
13072         from "rtx" to "rtx_insn *".
13074 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
13076         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
13077         bodies streamed in with -Q.
13078         * dumpfile.c (dump_files): Add lto-stream-out dump file.
13079         * dumpfile.h (tree_dump_index): Add lto_stream_out.
13080         * gimple-streamer-out.c: Include gimple-pretty-print.h
13081         (output_bb): Dump stmts streamed.
13082         * lto-section-out.c: Include print-tree.h
13083         (lto_begin_section): Dump sections created.
13084         (lto_output_decl_index): Dump decl encoded.
13085         * lto-streamer-out.c: Include print-tree.h
13086         (create_output_block): Dump output block created.
13087         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
13088         (output_function): Dump function output.
13089         (output_constructor): Dump constructor streamed.
13090         (write_global_stream): Output indexes encoded.
13091         (produce_asm_for_decls): Dump streams encoded.
13092         * lto-streamer.c (streamer_dump_file): New global var.
13093         * lto-streamer.h (streamer_dump_file): Declare.
13094         * passes.c (ipa_write_summaries): Initialize streamer dump.
13095         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
13096         in.
13098 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
13100         PR target/86048
13101         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
13102         offsets for register save directives.  Emit a second batch of save
13103         directives, if need be, when the function accesses prior frames.
13105 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13107         * config/arc/fpu.md (fmasf4): Force operand to register.
13108         (fnmasf4): Likewise.
13110 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13112         * config/arc/arc-protos.h (arc_pad_return): Remove.
13113         * config/arc/arc.c (machine_function): Remove force_short_suffix
13114         and size_reason.
13115         (arc_print_operand): Adjust printing of '&'.
13116         (arc_verify_short): Remove conditional printing of short suffix.
13117         (arc_final_prescan_insn): Remove reference to size_reason.
13118         (pad_return): New function.
13119         (arc_reorg): Call pad_return.
13120         (arc_pad_return): Remove.
13121         (arc_init_machine_status): Remove reference to force_short_suffix.
13122         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
13123         (attr length): When attribute iscompact is true force to 2
13124         regardless; in the case of maybe check if we want to force the
13125         instruction to have 4 bytes length.
13126         (nopv): Change it to generate 4 byte long nop as well.
13127         (blockage): New pattern.
13128         (simple_return): Remove call to arc_pad_return.
13129         (p_return_i): Likewise.
13131 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13133         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
13135 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13137         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
13138         ARC cores.
13140 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13142         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
13143         for ARC700 and ARCv2.
13145 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
13147         PR target/86076
13148         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
13149         operands[2] instead of operands[1].
13152 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
13154         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
13155         case, check whether the outer register overlaps an unallocatable
13156         register, not just whether it fits the required class.
13158 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
13160         * poly-int.h (can_div_trunc_p): Add new overload in which all values
13161         are poly_ints.
13162         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
13163         (memrefs_conflict_p): Likewise.
13164         (init_alias_analysis): Likewise.
13165         * cfgexpand.c (expand_debug_expr): Likewise.
13166         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
13167         * cse.c (fold_rtx): Likewise.
13168         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
13169         * expr.c (emit_block_move_hints): Likewise.
13170         (clear_storage_hints, push_block, emit_push_insn): Likewise.
13171         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
13172         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
13173         (emit_group_store): Likewise.
13174         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
13175         to read the PRE/POST_MODIFY increment.
13176         * calls.c (store_one_arg): Use strip_offset.
13177         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
13178         poly_int_rtx_p.
13179         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
13180         by a VEC_SELECT.
13181         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
13182         (simplify_binary_operation_1): Extend CONST_INT handling to
13183         poly_int_rtx_p.
13184         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
13185         than a HOST_WIDE_INT.
13186         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
13187         poly_int64.
13188         (adjust_mems, add_stores): Update accodingly.
13189         (vt_canonicalize_addr): Track polynomial offsets.
13190         (emit_note_insn_var_location): Likewise.
13191         (vt_add_function_parameter): Likewise.
13192         (vt_initialize): Likewise.
13194 2018-06-12  Jeff Law  <law@redhat.com>
13196         * config.gcc (alpha*-*-freebsd*): Remove.
13197         * config/alpha/freebsd.h: Remove.
13199 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
13201         PR other/69968
13202         * spellcheck-tree.c (levenshtein_distance): Rename to...
13203         (get_edit_distance): ...this, and update for underlying renaming.
13204         * spellcheck-tree.h (levenshtein_distance): Rename to...
13205         (get_edit_distance): ...this.
13206         * spellcheck.c (levenshtein_distance): Rename to...
13207         (get_edit_distance): ...this.  Convert from Levenshtein distance
13208         to Damerau-Levenshtein distance by supporting transpositions of
13209         adjacent characters.  Rename "v1" to "v_next" and "v0" to
13210         "v_one_ago".
13211         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
13212         (selftest::test_edit_distance_unit_test_oneway): ...this, and
13213         update for underlying renaming.
13214         (selftest::levenshtein_distance_unit_test): Rename to...
13215         (selftest::test_get_edit_distance_unit): ...this, and update for
13216         underlying renaming.
13217         (selftest::test_find_closest_string): Add example from PR 69968
13218         where transposition helps
13219         (selftest::test_metric_conditions): Update for renaming.
13220         (selftest::test_metric_conditions): Likewise.
13221         (selftest::spellcheck_c_tests): Likewise.
13222         * spellcheck.h (levenshtein_distance): Rename both overloads to...
13223         (get_edit_distance): ...this.
13224         (best_match::consider): Update for renaming.
13226 2018-06-12  Martin Sebor  <msebor@redhat.com>
13228         PR tree-optimization/85259
13229         * builtins.c (compute_objsize): Handle constant offsets.
13230         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
13231         true iff a warning has been issued.
13232         * gimple.h (gimple_nonartificial_location): New function.
13233         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
13234         gimple_nonartificial_location and handle -Wno-system-headers.
13235         (handle_builtin_stxncpy): Same.
13237 2018-06-12  Martin Sebor  <msebor@redhat.com>
13239         PR c/85931
13240         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
13242 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
13244         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
13245         BUILTIN_VEC_XST entries for pointer to double and long long.
13247 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
13249         PR target/85990
13250         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
13251         Update comments.
13252         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
13253         Likewise.
13255 2018-06-12  Martin Liska  <mliska@suse.cz>
13257         * doc/options.texi: Document IntegerRange.
13259 2018-06-12  Martin Liska  <mliska@suse.cz>
13261         * config/i386/i386.opt: Make MPX-related options as Deprecated.
13262         * opt-functions.awk: Handle Deprecated flag.
13263         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
13264         and report error.
13265         (read_cmdline_option): Report warning for a deprecated option.
13266         * opts.h (struct cl_option): Add new field cl_deprecated.
13267         (CL_ERR_DEPRECATED): New.
13269 2018-06-12  Martin Liska  <mliska@suse.cz>
13271         * doc/options.texi: Document Deprecated option flag.
13273 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13275         * config/arc/arc-arch.h (arc_extras): New enum.
13276         (arc_cpu_t):Add field extra.
13277         (arc_cpu_types): Consider the extras.
13278         * config/arc/arc-cpus.def: Add extras info.
13279         * config/arc/arc-opts.h (processor_type): Consider extra field.
13280         * config/arc/arc.c (arc_override_options): Handle extra field.
13282 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
13284         * config/arc/arc-arch.h: Update ARC_OPTX macro.
13285         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
13286         field.
13287         * config/arc/arc.c (arc_init): Update pic warning.
13288         (irq_range): Update irq range parsing warnings.
13289         (arc_override_options): Update various warning messages.
13290         (arc_handle_aux_attribute): Likewise.
13292 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
13294         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
13296 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13298         * doc/sourcebuild.texi: Document usage of line number 0 in verify
13299         compiler messages directives.
13301 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
13303         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
13304         * config/mips/mips-tables.opt: Regenerate.
13305         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
13306         mips64r6.
13307         * doc/invoke.texi: Document -march=i6500.
13309 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
13311         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
13312         (i6400_gpmul): Add cpu_unit.
13313         (i6400_gpdiv): Likewise.
13314         (i6400_msa_add_d): Update reservations.
13315         (i6400_msa_int_add) Likewise.
13316         (i6400_msa_short_logic3) Likewise.
13317         (i6400_msa_short_logic2) Likewise.
13318         (i6400_msa_short_logic) Likewise.
13319         (i6400_msa_move) Likewise.
13320         (i6400_msa_cmp) Likewise.
13321         (i6400_msa_short_float2) Likewise.
13322         (i6400_msa_div_d) Likewise.
13323         (i6400_msa_long_logic1) Likewise.
13324         (i6400_msa_long_logic2) Likewise.
13325         (i6400_msa_mult) Likewise.
13326         (i6400_msa_long_float2) Likewise.
13327         (i6400_msa_long_float4) Likewise.
13328         (i6400_msa_long_float5) Likewise.
13329         (i6400_msa_long_float8) Likewise.
13330         (i6400_fpu_fadd): Include frint type.
13331         (i6400_fpu_store): New define_insn_reservation.
13332         (i6400_fpu_load): Likewise.
13333         (i6400_fpu_move): Likewise.
13334         (i6400_fpu_fcmp): Likewise.
13335         (i6400_fpu_fmadd): Likewise.
13336         (i6400_int_mult): Include imul3nc type and update reservation.
13337         (i6400_int_div): Include idiv3 type and update reservation.
13338         (i6400_int_load): Update to check type not move_type.
13339         (i6400_int_store): Likewise.
13340         (i6400_int_prefetch): Set zero latency.
13342 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
13344         * gcc.c: Document new %@{...} sequence.
13345         (LINK_COMMAND_SPEC): Use it for the -L switches.
13346         (cpp_unique_options): Use it for the -I switches.
13347         (at_file_argbuf): New global variable.
13348         (in_at_file): Likewise.
13349         (alloc_args): Create at_file_argbuf.
13350         (clear_args): Truncate at_file_argbuf.
13351         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
13352         (open_at_file): New function.
13353         (close_at_file): Likewise.
13354         (create_at_file): Delete.
13355         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
13356         <'o'>: Likewise.
13357         <'@'>: New case.
13358         (validate_switches_from_spec): Deal with %@{...} sequence.
13359         (validate_switches): Likewise.
13360         (driver::finalize): Call clear_args.
13362 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
13364         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
13366 2018-06-11  Martin Sebor  <msebor@redhat.com>
13368         * doc/invoke.texi (-Wall): List -Wc++17-compat.
13369         (Wno-class-memaccess): Add @opindex.
13370         (Wno-templates, Wno-multiple-inheritance): Same.
13371         (Wno-virtual-inheritance, Wno-namespaces): Same.
13372         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
13373         (Wno-format-overflow, Wno-format-truncation): Same.
13374         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
13375         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
13376         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
13377         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
13378         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
13379         (Wno-misspelled-isr): Same.
13381 2018-06-11  Martin Sebor  <msebor@redhat.com>
13383         * PR tree-optimization/86083
13384         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
13386 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
13388         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
13390 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
13392         PR target/85755
13393         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
13394         on the correct operand.
13395         (*movdi_internal64): Ditto.
13397 2018-06-11  Martin Liska  <mliska@suse.cz>
13399         PR tree-optimization/86089
13400         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
13402 2018-06-11  Julia Koval  <julia.koval@intel.com>
13404         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
13405         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
13406         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
13408 2018-06-11  Olivier Hainque  <hainque@adacore.com>
13410         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
13411         for Ada with strict dwarf2.
13413 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
13415         PR target/85755
13416         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
13417         addresses.
13419 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
13421         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
13423 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
13425         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
13426         TARGET_ELF.
13428 2018-06-08  Martin Liska  <mliska@suse.cz>
13430         * tree-cfg.h (debug_function): Fix argument type to match
13431         implementation.
13433 2018-06-08  Martin Liska  <mliska@suse.cz>
13435         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
13436         Remove usage of MPX-related (and removed) fields.
13437         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
13439 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
13441         * cfg.c (debug): Use TDF_NONE rather than 0.
13442         * cfghooks.c (debug): Likewise.
13443         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
13444         (struct dump_option_value_info): Convert to...
13445         (struct kv_pair): ...this template type.
13446         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
13447         rather than 0.
13448         (optinfo_verbosity_options): Likewise.
13449         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
13450         OPTGROUP_NONE.
13451         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
13452         than int for "optgroup_flags" param.
13453         (dump_generic_expr_loc): Use dump_flags_t rather than int for
13454         "dump_kind" param.
13455         (dump_dec): Likewise.
13456         (dump_finish): Use TDF_NONE rather than 0.
13457         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
13458         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
13459         than 0.  Update for change to option_ptr.
13460         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
13461         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
13462         0.  Update for changes to optinfo_verbosity_options and
13463         optgroup_options.
13464         (opt_info_switch_p): Convert optgroup_flags from int to
13465         optgroup_flags_t.
13466         (dump_basic_block): Use dump_flags_t rather than int
13467         for "dump_kind" param.
13468         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
13469         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
13470         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
13471         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
13472         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
13473         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
13474         TDF_NONE): Convert from macros to...
13475         (enum dump_flag): ...this new enum.
13476         (dump_flags_t): Update to use enum.
13477         (operator|, operator&, operator~, operator|=, operator&=):
13478         Implement for dump_flags_t.
13479         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
13480         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
13481         Convert from macros to...
13482         (enum optgroup_flag): ...this new enum.
13483         (optgroup_flags_t): New typedef.
13484         (operator|, operator|=): Implement for optgroup_flags_t.
13485         (struct dump_file_info): Convert field "alt_flags" to
13486         dump_flags_t.  Convert field "optgroup_flags" to
13487         optgroup_flags_t.
13488         (dump_basic_block): Use dump_flags_t rather than int for param.
13489         (dump_generic_expr_loc): Likewise.
13490         (dump_dec): Likewise.
13491         (dump_register): Convert param "optgroup_flags" to
13492         optgroup_flags_t.
13493         (opt_info_enable_passes): Likewise.
13494         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
13495         than 0.
13496         * gimple-pretty-print.c (debug): Likewise.
13497         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
13498         (merged_store_group::apply_stores): Likewise.
13499         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
13500         * gimple.c (verify_gimple_pp): Likewise.
13501         * graphite-poly.c (print_pbb_body): Likewise.
13502         * passes.c (pass_manager::register_one_dump_file): Convert
13503         local "optgroup_flags" to optgroup_flags_t.
13504         * print-tree.c (print_node): Use TDF_NONE rather than 0.
13505         (debug): Likewise.
13506         (debug_body): Likewise.
13507         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
13508         to optgroup_flags_t.
13509         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
13510         than 0.
13511         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
13512         (convert_mult_to_fma): Likewise.
13513         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
13514         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
13515         * tree-vect-data-refs.c (dump_lower_bound): Convert param
13516         "dump_kind" to dump_flags_t.
13518 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
13520         * config/rs6000/rs6000.c (min, max): Delete.
13522 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
13524         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
13525         -mabi=spe and -mabi=no-spe.
13527 2018-06-08  Martin Liska  <mliska@suse.cz>
13529         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
13530         where we expect an existing summary.
13532 2018-06-08  Martin Liska  <mliska@suse.cz>
13534         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
13535         * ipa-inline.h (estimate_edge_growth): Likewise.
13537 2018-06-08  Martin Liska  <mliska@suse.cz>
13539         * cgraph.c (function_version_hasher::hash): Use
13540         cgraph_node::get_uid ().
13541         (function_version_hasher::equal):
13542         * cgraph.h (cgraph_node::get_uid): New method.
13543         * ipa-inline.c (update_caller_keys): Use
13544         cgraph_node::get_uid ().
13545         (update_callee_keys): Likewise.
13546         * ipa-utils.c (searchc): Likewise.
13547         (ipa_reduced_postorder): Likewise.
13548         * lto-cgraph.c (input_node): Likewise.
13549         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
13550         * symbol-summary.h (symtab_insertion): Likewise.
13551         (symtab_removal): Likewise.
13552         (symtab_duplication): Likewise.
13553         * tree-pretty-print.c (dump_function_header): Likewise.
13554         * tree-sra.c (convert_callers_for_node): Likewise.
13556 2018-06-08  Martin Liska  <mliska@suse.cz>
13558         * cgraph.c (symbol_table::create_edge): Always assign a new
13559         unique number.
13560         (symbol_table::free_edge): Do not recycle numbers.
13561         * cgraph.h (cgraph_edge::get): New method.
13562         * symbol-summary.h (symtab_removal): Use it.
13563         (symtab_duplication): Likewise.
13564         (call_summary::hashable_uid): Remove.
13566 2018-06-08  Martin Liska  <mliska@suse.cz>
13568         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
13569         (initialize_growth_caches): Remove.
13570         (free_growth_caches): Likewise.
13571         (do_estimate_edge_time): Use edge_growth_cache.
13572         (do_estimate_edge_size): Likewise.
13573         (do_estimate_edge_hints): Likewise.
13574         * ipa-inline.c (reset_edge_caches): Likewise.
13575         (recursive_inlining): Likewise.
13576         (inline_small_functions): Likewise.
13577         * ipa-inline.h (initialize_growth_caches): Remove.
13578         (estimate_edge_size): Likewise.
13579         (estimate_edge_time): Likewise.
13580         (estimate_edge_hints): Likewise.
13581         (reset_edge_growth_cache): Likewise.
13582         * symbol-summary.h (call_summary::remove): New method.
13584 2018-06-08  Martin Liska  <mliska@suse.cz>
13586         * ipa-cp.c (class edge_clone_summary): New summary.
13587         (grow_edge_clone_vectors): Remove.
13588         (ipcp_edge_duplication_hook): Remove.
13589         (class edge_clone_summary_t): New call_summary class.
13590         (ipcp_edge_removal_hook): Remove.
13591         (edge_clone_summary_t::duplicate): New function.
13592         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
13593         (create_specialized_node): Likewise.
13594         (ipcp_driver): Initialize edge_clone_summaries and do not
13595         register hooks.
13597 2018-06-08  Martin Liska  <mliska@suse.cz>
13599         * symbol-summary.h (get): New function.
13600         (call_summary::m_initialize_when_cloning): New class member.
13602 2018-06-08  Martin Liska  <mliska@suse.cz>
13604         * cgraph.c (cgraph_node::remove): Do not recycle uid.
13605         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
13606         (symbol_table::allocate_cgraph_symbol): Do not set uid.
13607         * passes.c (uid_hash_t): Record removed_nodes by their uids.
13608         (remove_cgraph_node_from_order): Use the removed_nodes set.
13609         (do_per_function_toporder): Likwise.
13610         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
13611         instead of summary_uid.
13612         (symtab_removal): Likewise.
13613         (symtab_duplication): Likewise.
13615 2018-06-08  Martin Liska  <mliska@suse.cz>
13617         * ipa-cp.c (ipcp_store_bits_results): Use
13618         ipcp_transformation_sum.
13619         (ipcp_store_vr_results): Likewise.
13620         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
13621         to ...
13622         (ipcp_transformation_initialize): ... this.
13623         (ipa_set_node_agg_value_chain):
13624         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
13625         (write_ipcp_transformation_info): Likewise.
13626         (read_ipcp_transformation_info): Likewise.
13627         (ipcp_update_bits): Likewise.
13628         (ipcp_update_vr): Likewise.
13629         (ipcp_transform_function): Likewise.
13630         * ipa-prop.h: Rename ipcp_transformation_summary to
13631         ipcp_transformation.
13632         (class ipcp_transformation_t): New function summary.
13633         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
13634         (ipa_get_agg_replacements_for_node): Likewise.
13636 2018-06-08  Martin Liska  <mliska@suse.cz>
13638         * ipa-pure-const.c (struct funct_state_d): Do it class instead
13639         of struct.
13640         (class funct_state_summary_t): New function_summary class.
13641         (has_function_state): Remove.
13642         (get_function_state): Likewise.
13643         (set_function_state): Likewise.
13644         (add_new_function): Likewise.
13645         (funct_state_summary_t::insert): New function.
13646         (duplicate_node_data): Remove.
13647         (remove_node_data): Remove.
13648         (funct_state_summary_t::duplicate): New function.
13649         (register_hooks): Create new funct_state_summaries.
13650         (pure_const_generate_summary): Use it.
13651         (pure_const_write_summary): Likewise.
13652         (pure_const_read_summary): Likewise.
13653         (propagate_pure_const): Likewise.
13654         (propagate_nothrow): Likewise.
13655         (dump_malloc_lattice): Likewise.
13656         (propagate_malloc): Likewise.
13657         (execute): Do not register hooks, just remove summary
13658         instead.
13659         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
13660         constructor.
13662 2018-06-08  Martin Liska  <mliska@suse.cz>
13664         * ipa-reference.c (remove_node_data): Remove.
13665         (duplicate_node_data): Likewise.
13666         (class ipa_ref_var_info_summary_t): New class.
13667         (class ipa_ref_opt_summary_t): Likewise.
13668         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
13669         (get_reference_optimization_summary): Use
13670         ipa_ref_opt_sum_summaries.
13671         (set_reference_vars_info): Remove.
13672         (set_reference_optimization_summary): Likewise.
13673         (ipa_init): Create summaries.
13674         (init_function_info): Use function summary.
13675         (ipa_ref_opt_summary_t::duplicate): New function.
13676         (ipa_ref_opt_summary_t::remove): New function.
13677         (get_read_write_all_from_node): Fix GNU coding style.
13678         (propagate): Use function summary.
13679         (write_node_summary_p): Fix GNU coding style.
13680         (stream_out_bitmap): Likewise.
13681         (ipa_reference_read_optimization_summary): Use function summary.
13682         (ipa_reference_c_finalize): Do not release hooks.
13684 2018-06-08  Martin Liska  <mliska@suse.cz>
13686         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
13687         (analyze_function_body): Extract multiple calls of get_create.
13688         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
13689         * ipa-inline.c (recursive_inlining): Use ::get method.
13690         * ipa-inline.h (estimate_edge_growth): Likewise.
13692 2018-06-08  Martin Liska  <mliska@suse.cz>
13694         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
13695         HSA_INVALID.
13696         (hsa_function_summary::hsa_function_summary): Use the new enum
13697         value.
13698         (hsa_gpu_implementation_p): Use hsa_summaries::get.
13699         * hsa-gen.c (hsa_get_host_function): Likewise.
13700         (get_brig_function_name): Likewise.
13701         * ipa-hsa.c (process_hsa_functions): Likewise.
13702         (ipa_hsa_write_summary): Likewise.
13703         * symbol-summary.h (symtab_duplication): Use ::get function/
13704         (get): New function.
13706 2018-06-08  Martin Liska  <mliska@suse.cz>
13708         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
13709         of get.
13710         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
13711         (hsa_register_kernel): Likewise.
13712         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
13713         * hsa-gen.c (hsa_get_host_function): Likewise.
13714         (get_brig_function_name): Likewise.
13715         (generate_hsa): Likewise.
13716         (pass_gen_hsail::execute): Likewise.
13717         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
13718         (devirtualization_time_bonus): Likewise.
13719         (ipcp_propagate_stage): Likewise.
13720         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
13721         (edge_set_predicate): Likewise.
13722         (evaluate_conditions_for_known_args): Likewise.
13723         (evaluate_properties_for_edge): Likewise.
13724         (ipa_fn_summary::reset): Likewise.
13725         (ipa_fn_summary_t::duplicate): Likewise.
13726         (dump_ipa_call_summary): Likewise.
13727         (ipa_dump_fn_summary): Likewise.
13728         (analyze_function_body): Likewise.
13729         (compute_fn_summary): Likewise.
13730         (estimate_edge_devirt_benefit): Likewise.
13731         (estimate_edge_size_and_time): Likewise.
13732         (estimate_calls_size_and_time): Likewise.
13733         (estimate_node_size_and_time): Likewise.
13734         (inline_update_callee_summaries): Likewise.
13735         (remap_edge_change_prob): Likewise.
13736         (remap_edge_summaries): Likewise.
13737         (ipa_merge_fn_summary_after_inlining): Likewise.
13738         (ipa_update_overall_fn_summary): Likewise.
13739         (read_ipa_call_summary): Likewise.
13740         (inline_read_section): Likewise.
13741         (write_ipa_call_summary): Likewise.
13742         (ipa_fn_summary_write): Likewise.
13743         (ipa_free_fn_summary): Likewise.
13744         * ipa-hsa.c (process_hsa_functions): Likewise.
13745         (ipa_hsa_write_summary): Likewise.
13746         (ipa_hsa_read_section): Likewise.
13747         * ipa-icf.c (sem_function::merge): Likewise.
13748         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
13749         (do_estimate_edge_time): Likewise.
13750         (estimate_size_after_inlining): Likewise.
13751         (estimate_growth): Likewise.
13752         (growth_likely_positive): Likewise.
13753         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
13754         (inline_call): Likewise.
13755         * ipa-inline.c (caller_growth_limits): Likewise.
13756         (can_inline_edge_p): Likewise.
13757         (can_inline_edge_by_limits_p): Likewise.
13758         (compute_uninlined_call_time): Likewise.
13759         (compute_inlined_call_time): Likewise.
13760         (want_inline_small_function_p): Likewise.
13761         (edge_badness): Likewise.
13762         (update_caller_keys): Likewise.
13763         (update_callee_keys): Likewise.
13764         (recursive_inlining): Likewise.
13765         (inline_small_functions): Likewise.
13766         (inline_to_all_callers_1): Likewise.
13767         (dump_overall_stats): Likewise.
13768         (early_inline_small_functions): Likewise.
13769         (early_inliner): Likewise.
13770         * ipa-inline.h (estimate_edge_growth): Likewise.
13771         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
13772         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
13773         * ipa-prop.h (IPA_NODE_REF): Likewise.
13774         (IPA_EDGE_REF): Likewise.
13775         * ipa-pure-const.c (malloc_candidate_p): Likewise.
13776         (propagate_malloc): Likewise.
13777         * ipa-split.c (execute_split_functions): Likewise.
13778         * symbol-summary.h: Rename get to get_create.
13779         (get): Likewise.
13780         (get_create): Likewise.
13781         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
13783 2018-06-08  Martin Liska  <mliska@suse.cz>
13785         * symbol-summary.h (release): Move definition out of class
13786         declaration.
13787         (symtab_removal): Likewise.
13788         (symtab_duplication): Likewise.
13790 2018-06-08  Martin Liska  <mliska@suse.cz>
13792         * symbol-summary.h (function_summary): Move constructor
13793         implementation out of class declaration.
13794         (release): Likewise.
13795         (symtab_insertion): Likewise.
13796         (symtab_removal): Likewise.
13797         (symtab_duplication): Likewise.
13798         (get): Likewise.
13800 2018-06-08  Martin Liska  <mliska@suse.cz>
13802         * Makefile.in: Remove support for MPX (macros, related functions,
13803         fields in cgraph_node, ...).
13804         * builtin-types.def (BT_BND): Likewise.
13805         (BT_FN_BND_CONST_PTR): Likewise.
13806         (BT_FN_CONST_PTR_BND): Likewise.
13807         (BT_FN_VOID_PTR_BND): Likewise.
13808         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
13809         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
13810         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
13811         (expand_builtin_mempcpy_with_bounds): Likewise.
13812         (expand_builtin_memset_with_bounds): Likewise.
13813         (expand_builtin_memset_args): Likewise.
13814         (std_expand_builtin_va_start): Likewise.
13815         (expand_builtin): Likewise.
13816         (expand_builtin_with_bounds): Likewise.
13817         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
13818         (DEF_LIB_BUILTIN_CHKP): Likewise.
13819         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
13820         (DEF_CHKP_BUILTIN): Likewise.
13821         (BUILT_IN_MEMCPY): Likewise.
13822         (BUILT_IN_MEMMOVE): Likewise.
13823         (BUILT_IN_MEMPCPY): Likewise.
13824         (BUILT_IN_MEMSET): Likewise.
13825         (BUILT_IN_STPCPY): Likewise.
13826         (BUILT_IN_STRCAT): Likewise.
13827         (BUILT_IN_STRCHR): Likewise.
13828         (BUILT_IN_STRCPY): Likewise.
13829         (BUILT_IN_STRLEN): Likewise.
13830         (BUILT_IN_MEMCPY_CHK): Likewise.
13831         (BUILT_IN_MEMMOVE_CHK): Likewise.
13832         (BUILT_IN_MEMPCPY_CHK): Likewise.
13833         (BUILT_IN_MEMSET_CHK): Likewise.
13834         (BUILT_IN_STPCPY_CHK): Likewise.
13835         (BUILT_IN_STRCAT_CHK): Likewise.
13836         (BUILT_IN_STRCPY_CHK): Likewise.
13837         * calls.c (store_bounds): Likewise.
13838         (emit_call_1): Likewise.
13839         (special_function_p): Likewise.
13840         (maybe_warn_nonstring_arg): Likewise.
13841         (initialize_argument_information): Likewise.
13842         (finalize_must_preallocate): Likewise.
13843         (compute_argument_addresses): Likewise.
13844         (expand_call): Likewise.
13845         * cfgexpand.c (expand_call_stmt): Likewise.
13846         (expand_return): Likewise.
13847         (expand_gimple_stmt_1): Likewise.
13848         (pass_expand::execute): Likewise.
13849         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
13850         (cgraph_node::remove): Likewise.
13851         (cgraph_node::dump): Likewise.
13852         (cgraph_node::verify_node): Likewise.
13853         * cgraph.h (chkp_function_instrumented_p): Likewise.
13854         (symtab_node::get_alias_target): Likewise.
13855         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
13856         (cgraph_local_p): Likewise.
13857         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
13858         (cgraph_edge::rebuild_references): Likewise.
13859         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
13860         (walk_polymorphic_call_targets): Likewise.
13861         (cgraph_node::expand_thunk): Likewise.
13862         (symbol_table::output_weakrefs): Likewise.
13863         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
13864         (ix86_handle_option): Likewise.
13865         * config/i386/constraints.md: Likewise.
13866         * config/i386/i386-builtin-types.def (BND): Likewise.
13867         (VOID): Likewise.
13868         (PVOID): Likewise.
13869         (ULONG): Likewise.
13870         * config/i386/i386-builtin.def (BDESC_END): Likewise.
13871         (BDESC_FIRST): Likewise.
13872         (BDESC): Likewise.
13873         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
13874         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
13875         * config/i386/i386.c (enum reg_class): Likewise.
13876         (ix86_target_string): Likewise.
13877         (ix86_option_override_internal): Likewise.
13878         (ix86_conditional_register_usage): Likewise.
13879         (ix86_valid_target_attribute_inner_p): Likewise.
13880         (ix86_set_indirect_branch_type): Likewise.
13881         (ix86_set_current_function): Likewise.
13882         (ix86_function_arg_regno_p): Likewise.
13883         (init_cumulative_args): Likewise.
13884         (ix86_function_arg_advance): Likewise.
13885         (ix86_function_arg): Likewise.
13886         (ix86_pass_by_reference): Likewise.
13887         (ix86_function_value_regno_p): Likewise.
13888         (ix86_function_value_1): Likewise.
13889         (ix86_function_value_bounds): Likewise.
13890         (ix86_return_in_memory): Likewise.
13891         (ix86_setup_incoming_vararg_bounds): Likewise.
13892         (ix86_va_start): Likewise.
13893         (indirect_thunk_need_prefix): Likewise.
13894         (print_reg): Likewise.
13895         (ix86_print_operand): Likewise.
13896         (ix86_expand_call): Likewise.
13897         (ix86_output_function_return): Likewise.
13898         (reg_encoded_number): Likewise.
13899         (BDESC_VERIFYS): Likewise.
13900         (ix86_init_mpx_builtins): Likewise.
13901         (ix86_init_builtins): Likewise.
13902         (ix86_emit_cmove): Likewise.
13903         (ix86_emit_move_max): Likewise.
13904         (ix86_expand_builtin): Likewise.
13905         (ix86_builtin_mpx_function): Likewise.
13906         (ix86_get_arg_address_for_bt): Likewise.
13907         (ix86_load_bounds): Likewise.
13908         (ix86_store_bounds): Likewise.
13909         (ix86_load_returned_bounds): Likewise.
13910         (ix86_store_returned_bounds): Likewise.
13911         (ix86_class_likely_spilled_p): Likewise.
13912         (ix86_hard_regno_mode_ok): Likewise.
13913         (x86_order_regs_for_local_alloc): Likewise.
13914         (ix86_mitigate_rop): Likewise.
13915         (ix86_bnd_prefixed_insn_p): Likewise.
13916         (ix86_mpx_bound_mode): Likewise.
13917         (ix86_make_bounds_constant): Likewise.
13918         (ix86_initialize_bounds): Likewise.
13919         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
13920         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
13921         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
13922         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
13923         (TARGET_CHKP_BOUND_MODE): Likewise.
13924         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
13925         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
13926         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
13927         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
13928         * config/i386/i386.h (TARGET_MPX): Likewise.
13929         (TARGET_MPX_P): Likewise.
13930         (VALID_BND_REG_MODE): Likewise.
13931         (FIRST_BND_REG): Likewise.
13932         (LAST_BND_REG): Likewise.
13933         (enum reg_class): Likewise.
13934         (BND_REG_P): Likewise.
13935         (BND_REGNO_P): Likewise.
13936         (BNDmode): Likewise.
13937         (ADJUST_INSN_LENGTH): Likewise.
13938         * config/i386/i386.md: Likewise.
13939         * config/i386/i386.opt: Likewise.
13940         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
13941         (defined): Likewise.
13942         (LINK_MPX): Likewise.
13943         (MPX_SPEC): Likewise.
13944         (LIBMPX_SPEC): Likewise.
13945         (LIBMPXWRAPPERS_SPEC): Likewise.
13946         (CHKP_SPEC): Likewise.
13947         * config/i386/predicates.md: Likewise.
13948         * dbxout.c (dbxout_type): Likewise.
13949         * doc/extend.texi: Likewise.
13950         * doc/invoke.texi: Likewise.
13951         * doc/md.texi: Likewise.
13952         * doc/tm.texi: Likewise.
13953         * doc/tm.texi.in: Likewise.
13954         * dwarf2out.c (is_base_type): Likewise.
13955         (gen_formal_types_die): Likewise.
13956         (gen_subprogram_die): Likewise.
13957         (gen_type_die_with_usage): Likewise.
13958         (gen_decl_die): Likewise.
13959         (dwarf2out_late_global_decl): Likewise.
13960         * expr.c (expand_assignment): Likewise.
13961         (emit_storent_insn): Likewise.
13962         (store_expr_with_bounds): Likewise.
13963         (store_expr): Likewise.
13964         (expand_expr_real_1): Likewise.
13965         * expr.h (store_expr_with_bounds): Likewise.
13966         * function.c (use_register_for_decl): Likewise.
13967         (struct bounds_parm_data): Likewise.
13968         (assign_parms_augmented_arg_list): Likewise.
13969         (assign_parm_find_entry_rtl): Likewise.
13970         (assign_parm_is_stack_parm): Likewise.
13971         (assign_parm_load_bounds): Likewise.
13972         (assign_bounds): Likewise.
13973         (assign_parms): Likewise.
13974         (expand_function_start): Likewise.
13975         * gcc.c (CHKP_SPEC): Likewise.
13976         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
13977         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
13978         (wrestrict_dom_walker::check_call): Likewise.
13979         * gimple.c (gimple_build_call_from_tree): Likewise.
13980         * gimple.h (enum gf_mask): Likewise.
13981         (gimple_call_with_bounds_p): Likewise.
13982         (gimple_call_set_with_bounds): Likewise.
13983         * gimplify.c (gimplify_init_constructor): Likewise.
13984         * ipa-cp.c (initialize_node_lattices): Likewise.
13985         (propagate_constants_across_call): Likewise.
13986         (find_more_scalar_values_for_callers_subset): Likewise.
13987         * ipa-hsa.c (process_hsa_functions): Likewise.
13988         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
13989         * ipa-icf.c (sem_function::merge): Likewise.
13990         * ipa-inline.c (early_inliner): Likewise.
13991         * ipa-pure-const.c (warn_function_noreturn): Likewise.
13992         (warn_function_cold): Likewise.
13993         (propagate_pure_const): Likewise.
13994         * ipa-ref.h (enum GTY): Likewise.
13995         * ipa-split.c (find_retbnd): Likewise.
13996         (consider_split): Likewise.
13997         (split_function): Likewise.
13998         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
13999         * ipa.c (walk_polymorphic_call_targets): Likewise.
14000         (symbol_table::remove_unreachable_nodes): Likewise.
14001         (process_references): Likewise.
14002         (cgraph_build_static_cdtor_1): Likewise.
14003         * lto-cgraph.c (lto_output_node): Likewise.
14004         (output_refs): Likewise.
14005         (compute_ltrans_boundary): Likewise.
14006         (input_overwrite_node): Likewise.
14007         (input_node): Likewise.
14008         (input_cgraph_1): Likewise.
14009         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
14010         * passes.c (pass_manager::execute_early_local_passes): Likewise.
14011         (class pass_chkp_instrumentation_passes): Likewise.
14012         (make_pass_chkp_instrumentation_passes): Likewise.
14013         * passes.def: Likewise.
14014         * rtl.h (struct GTY): Likewise.
14015         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
14016         * stor-layout.c (layout_type): Likewise.
14017         * symtab.c: Likewise.
14018         * target.def: Likewise.
14019         * targhooks.c (default_chkp_bound_type): Likewise.
14020         (default_chkp_bound_mode): Likewise.
14021         (default_builtin_chkp_function): Likewise.
14022         (default_chkp_function_value_bounds): Likewise.
14023         (default_chkp_make_bounds_constant): Likewise.
14024         (default_chkp_initialize_bounds): Likewise.
14025         * targhooks.h (default_chkp_bound_type): Likewise.
14026         (default_chkp_bound_mode): Likewise.
14027         (default_builtin_chkp_function): Likewise.
14028         (default_chkp_function_value_bounds): Likewise.
14029         (default_chkp_make_bounds_constant): Likewise.
14030         (default_chkp_initialize_bounds): Likewise.
14031         * toplev.c (compile_file): Likewise.
14032         (process_options): Likewise.
14033         * tree-core.h (DEF_BUILTIN): Likewise.
14034         (DEF_BUILTIN_CHKP): Likewise.
14035         * tree-inline.c (declare_return_variable): Likewise.
14036         (remap_gimple_stmt): Likewise.
14037         (copy_bb): Likewise.
14038         (initialize_inlined_parameters): Likewise.
14039         (expand_call_inline): Likewise.
14040         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
14041         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
14042         (make_pass_ipa_chkp_produce_thunks): Likewise.
14043         (make_pass_chkp): Likewise.
14044         (make_pass_chkp_opt): Likewise.
14045         (make_pass_chkp_instrumentation_passes): Likewise.
14046         * tree-pretty-print.c (dump_generic_node): Likewise.
14047         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
14048         * tree-ssa-dce.c (propagate_necessity): Likewise.
14049         (eliminate_unnecessary_stmts): Likewise.
14050         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
14051         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
14052         * tree-ssa-sccvn.h: Likewise.
14053         * tree-ssa-strlen.c (get_string_length): Likewise.
14054         (valid_builtin_call): Likewise.
14055         (adjust_last_stmt): Likewise.
14056         (handle_builtin_strchr): Likewise.
14057         (handle_builtin_strcpy): Likewise.
14058         (handle_builtin_stxncpy): Likewise.
14059         (handle_builtin_memcpy): Likewise.
14060         (handle_builtin_strcat): Likewise.
14061         (strlen_check_and_optimize_stmt): Likewise.
14062         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
14063         * tree-streamer-in.c: Likewise.
14064         * tree-streamer.c (record_common_node): Likewise.
14065         * tree.c (tree_code_size): Likewise.
14066         (wide_int_to_tree_1): Likewise.
14067         (type_contains_placeholder_1): Likewise.
14068         (build_common_tree_nodes): Likewise.
14069         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
14070         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
14071         (POINTER_BOUNDS_P): Likewise.
14072         (BOUNDED_TYPE_P): Likewise.
14073         (BOUNDED_P): Likewise.
14074         (CALL_WITH_BOUNDS_P): Likewise.
14075         (pointer_bounds_type_node): Likewise.
14076         * value-prof.c (gimple_ic): Likewise.
14077         * var-tracking.c (vt_add_function_parameters): Likewise.
14078         * varasm.c (make_decl_rtl): Likewise.
14079         (assemble_start_function): Likewise.
14080         (output_constant): Likewise.
14081         (maybe_assemble_visibility): Likewise.
14082         * varpool.c (ctor_for_folding): Likewise.
14083         * chkp-builtins.def: Remove.
14084         * ipa-chkp.c: Remove.
14085         * ipa-chkp.h: Remove.
14086         * rtl-chkp.c: Remove.
14087         * rtl-chkp.h: Remove.
14088         * tree-chkp-opt.c: Remove.
14089         * tree-chkp.c: Remove.
14090         * tree-chkp.h: Remove.
14092 2018-06-07  Carl Love  <cel@us.ibm.com>
14094         * config/rs6000/vsx.md (vextract_fp_from_shorth,
14095         vextract_fp_from_shortl): Add BE support.
14097 2018-06-07  Paul Koning  <ni1d@arrl.net>
14099         * compare-elim.c (try_merge_compare): Don't merge compare if
14100         address contains a side effect.
14101         (try_eliminate_compare): Likewise.
14103 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
14105         * config.gcc: Support "tremont".
14106         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
14107         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14108         PROCESSOR_TREMONT.
14109         * config/i386/i386.c (m_TREMONT): Define.
14110         (processor_target_table): Add "tremont".
14111         (PTA_TREMONT): Define.
14112         (ix86_lea_outperforms): Add TARGET_TREMONT.
14113         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
14114         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
14115         and M_INTEL_GOLDMONT_PLUS.
14116         (fold_builtin_cpu): Add "tremont".
14117         (ix86_add_stmt_cost): Add TARGET_TREMONT.
14118         (ix86_option_override_internal): Add "tremont".
14119         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
14120         (processor_type): Add PROCESSOR_TREMONT.
14121         * config/i386/x86-tune.def: Add m_TREMONT.
14122         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
14124 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14126         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
14127         symbol defined for msp430i* devices to be lower case.
14129 2018-06-07  Richard Biener  <rguenther@suse.de>
14131         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
14132         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
14133         Properly wrap signed arithmetic if overflow wraps.
14135 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
14137         PR tree-optimization/69615
14138         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
14139         of a cast from a same precision integral SSA_NAME in a bb dominated
14140         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
14141         cast to utype if rhs2 has already a compatible type.
14143 2018-06-07  Richard Biener  <rguenther@suse.de>
14145         PR tree-optimization/85935
14146         * graphite-scop-detection.c (find_params_in_bb): Analyze
14147         condition operands with respect to the correct loop.  Assert
14148         the analysis doesn't fail.
14150 2018-06-04  Carl Love  <cel@us.ibm.com>
14152         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
14153         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
14154         as it is slightly cheaper.
14155         (first_match_or_eos_index_<mode>):
14156         Calculate index using natural element order.
14157         (first_match_index_<mode>):
14158         Calculate index using natural element order.
14159         (first_match_or_eos_index_<mode>):
14160         Calculate index using natural order.
14161         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
14162         for BE and LE modes.
14163         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
14164         P9V_BUILTIN_VCLZLSBB_V16QI.
14165         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
14166         specific.
14168 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14170         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
14171         indentation and line wrap for many prototypes.  Add missing
14172         @smallexample directives around block of prototypes for vec_xl and
14173         vec_xst.
14175 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
14177         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
14178         track if we pass or return IEEE 128-bit floating point.
14179         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
14180         C++ mangling that is compatible with GCC 8.1.
14181         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
14182         (init_cumulative_args): Note if we pass or return IEEE 128-bit
14183         floating point types.
14184         (rs6000_function_arg_advance_1): Likewise.
14185         (rs6000_mangle_type): Optionally generate mangled names that match
14186         what GCC 8.1 generated for IEEE 128-bit floating point types.
14187         (rs6000_globalize_decl_name): If we have an external function that
14188         passes or returns IEEE 128-bit types, generate a weak reference
14189         from the mangled name used in GCC 8.1 to the current mangled
14190         name.
14191         (rs6000_init_builtins): Make __ibm128 use the long double type if
14192         long double is IBM extended double.  Make __float128 use the long
14193         double type if long double is IEEE 128-bit.
14195         PR target/85657
14196         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
14197         macro for __ibm128 built-in functions.
14198         (PACK_IF): Add __ibm128 pack/unpack functions.
14199         (UNPACK_IF): Likewise.
14200         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
14201         enable long double built-in functions if long double is IEEE
14202         128-bit floating point.
14203         (rs6000_invalid_builtin): Update long double built-in function
14204         error message.
14205         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
14206         functions, adjust the built-in function to use the long double
14207         built-in function if __ibm128 and long double are the same type.
14208         * doc/extend.texi (PowerPC builtins): Update documention for
14209         __builtin_{,un}pack_longdouble.  Add documentation for
14210         __builtin_{,un}pack_ibm128.
14212 2018-06-06  Jim Wilson  <jimw@sifive.com>
14214         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
14215         (struct machine_function): New field interrupt_mode.
14216         (riscv_handle_type_attribute): New function.  Add forward declaration.
14217         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
14218         (riscv_expand_epilogue): Check interrupt_mode field.
14219         (riscv_set_current_function): Check interrupt attribute args and
14220         set interrupt_mode field.
14221         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
14222         (riscv_sret, riscv_uret): New.
14223         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
14224         new arguments to interrupt attribute.
14226 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
14228         PR target/63177
14229         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
14230         Don't handle -mcpu=power8 if -mpower9-vector is also used.
14232 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14234         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
14235         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
14236         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
14237         several redundant entries.
14239 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
14241         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
14242         type from "rtx" to "rtx_insn *".
14243         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
14244         for local "call_insn", removing cast.
14245         (ix86_expand_call): Likewise, introducing a "call_insn" local.
14247 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
14249         PR tree-optimization/86066
14250         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
14251         for BIT_INSERT_EXPR stores.
14253 2018-06-06  Richard Biener  <rguenther@suse.de>
14255         PR tree-optimization/86062
14256         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
14257         component refs ontop
14258         of to be offsetted base.
14260 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14262         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
14263         to be static and remove check on interrupt attribute name.
14265 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14267         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
14268         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
14270 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
14272         PR target/79924
14273         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
14274         second argument.
14275         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
14276         Remove second argument, change how error is called.
14277         (aarch64_layout_arg): Remove second argument from
14278         aarch64_err_no_fpadvsimd call.
14279         (aarch64_init_cumulative_args): Ditto.
14280         (aarch64_gimplify_va_arg_expr): Ditto.
14281         * config/aarch64/aarch64.md (mov<mode>): Ditto.
14283 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
14285         * config/i386/i386.md (simple_return_indirect_internal): New expander.
14286         (*simple_return_indirect_internal<mode>): Rename from
14287         simple_return_indirect_internal.  Use W mode iterator.
14288         (rstorssp): New expander.
14289         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
14290         (clrssbsy): New expander.
14291         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
14293 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14295         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
14296         __typeof__.
14297         (cmse_check_pointed_object): Likewise.
14299 2018-06-05  Martin Liska  <mliska@suse.cz>
14301         PR gcov-profile/47618
14302         * doc/invoke.texi: Document how -fprofile-dir format
14303         is extended.
14305 2018-06-05  Richard Biener  <rguenther@suse.de>
14307         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
14308         removal pretend DOM info isn't available so we do not update
14309         it and only remove edges, not dominated blocks.  Actually free
14310         DOM info in case we removed something.  Remove unreachable blocks.
14311         (mfb_keep_latches): Work with either DOM info or marked backedges.
14312         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
14313         first.  Mark backedges if DOM info isn't available.
14314         (Re-)compute DOM info after cleanup_control_flow_pre.
14316 2018-06-05  Richard Biener  <rguenther@suse.de>
14318         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
14319         (locus_discrim_hasher::hash): Adjust.
14320         (locus_discrim_hasher::equal): Likewise.
14321         (next_discriminator_for_locus): Work on line directly.
14322         (same_line_p): Pass in expanded locus1 as well.
14323         (assign_discriminators): Avoid redundant location expansions.
14325 2018-06-05  Richard Biener  <rguenther@suse.de>
14327         PR tree-optimization/86046
14328         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
14329         if required after clearing TREE_ADDRESSABLE.
14331 2018-06-05  Richard Biener  <rguenther@suse.de>
14333         PR tree-optimization/86047
14334         * tree-ssa-loop.c (for_each_index): Glob handling of all
14335         decls and constants and really handle all of them.
14337 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14339         PR target/81497
14340         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
14341         qualifier_void_pointer and qualifier_const_void_pointer.
14342         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
14343         (arm_init_builtins): Handle the above.
14344         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
14345         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
14346         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
14347         void intrinsics.
14349 2018-06-05  Martin Liska  <mliska@suse.cz>
14351         * auto-profile.c (read_autofdo_file): Do not use
14352         gcov_ctr_summary struct.
14353         (afdo_callsite_hot_enough_for_early_inline): Likewise.
14354         * coverage.c (struct counts_entry): Likewise.
14355         (read_counts_file): Read just single summary entry.
14356         (get_coverage_counts): Use gcov_summary struct.
14357         * coverage.h (get_coverage_counts): Likewise.
14358         * gcov-dump.c (dump_working_sets): Likewise.
14359         (tag_summary): Dump just single summary.
14360         * gcov-io.c (gcov_write_summary): Write just histogram
14361         summary.
14362         (gcov_read_summary): Read just single summary.
14363         (compute_working_sets): Use gcov_summary struct.
14364         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
14365         of GCOV_COUNTERS_SUMMABLE.
14366         (GCOV_COUNTERS_SUMMABLE): Remove.
14367         (GCOV_FIRST_VALUE_COUNTER): Replace with
14368         GCOV_COUNTER_V_INTERVAL.
14369         (struct gcov_ctr_summary): Remove.
14370         (struct gcov_summary): Directly use fields of former
14371         gcov_ctr_summary.
14372         (compute_working_sets): Use gcov_summary struct.
14373         * gcov.c (read_count_file): Do not use ctrs fields.
14374         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
14375         struct.
14376         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
14377         struct.
14378         * profile.c: Likewise.
14379         * profile.h: Likewise.
14381 2018-06-05  Martin Liska  <mliska@suse.cz>
14383         PR gcov-profile/84846
14384         * gcov.c (output_lines): Print working directory only
14385         in intermediate format.
14387 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
14389         * config/s390/s390-builtin-types.def: Add void function type.
14390         * config/s390/s390-builtins.def: Use the function type for the
14391         tbeginc builtin.
14393 2018-06-04  Jim Wilson  <jimw@sifive.com>
14395         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
14396         to int.
14397         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
14398         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
14399         handle EH_RETURN_DATA_REGNO registers properly.
14400         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
14401         (riscv_expand_epilogue): Update comment.  Change argument name and
14402         type.  Update code to use new name and type.  Pass new args to
14403         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
14404         EXCEPTION_RETURN.
14405         * config/riscv/riscv.md (NORMAL_RETURN): New.
14406         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
14407         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
14408         (eh_return): Call gen_eh_return_internal and emit barrier.
14409         (eh_return_internal): Call riscv_expand_epilogue.
14411 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
14413         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
14414         bit_insertion field and declare can_be_merged_into method.
14415         (merged_store_group::can_be_merged_into): New method.
14416         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
14417         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
14418         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
14420 2018-06-04  Richard Biener  <rguenther@suse.de>
14422         PR tree-optimization/85955
14423         * builtins.c (fold_builtin_sincos): Convert pointers to
14424         destination to appropriate type before dereferencing.
14426 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
14428         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
14430 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
14432         * expr.c (expand_expr_real_1): Force the operand into memory if
14433         its TYPE_MODE is BLKmode and if there is no integer mode for
14434         the number of bits being extracted.
14436 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
14438         PR target/85832
14439         PR target/86036
14440         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
14441         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
14443 2018-06-04  Richard Biener  <rguenther@suse.de>
14445         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
14446         (cleanup_tree_cfg_noloop): ... single caller.  Do
14447         start_recording_case_labels later.
14449 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
14451         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
14452         to _IMMINTRIN_H_INCLUDED.
14453         * config/i386/pconfigintrin.h: Ditto.
14454         * config/i386/waitpkgintrin.h: Ditto.
14455         * config/i386/immintrin.h: Add includes for sgxintrin.h,
14456         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
14457         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
14458         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
14459         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
14460         waitpkgintrin.h and cldemoteintrin.h.
14462 2018-06-04  Richard Biener  <rguenther@suse.de>
14464         PR tree-optimization/86038
14465         * tracer.c (find_best_successor): Check probability for
14466         being initialized, bail out if not.
14468 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
14470         PR target/86003
14471         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
14472         of bits to ignore when comparing architectures.
14474 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
14476         PR tree-optimization/69615
14477         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
14478         maximum or minimum of the type, try to merge it also as if
14479         range1 is + [-, x - 1] or + [x + 1, -].
14481         PR c++/86025
14482         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
14484 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
14486         PR tree-optimization/86034
14487         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
14488         the unsigned bitfield type in a bit insertion sequence if it does not
14489         have a larger precision than the bitfield size.
14490         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
14492 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
14494         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
14496 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
14498         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
14499         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
14500         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
14501         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
14503 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
14505         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
14506         Disable -fdelete-null-pointer-checks for ELF toolchain.
14508 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
14509             Kito Cheng  <kito.cheng@gmail.com>
14511         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
14512         (nds32le-*-*, nds32be-*-*): Integrate checking process.
14513         (nds32*-*-*): Add glibc and uclibc conditions.
14514         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
14515         (TARGET_EXCEPT_UNWIND_INFO): Define.
14516         * config/nds32/elf.h: New file.
14517         * config/nds32/linux.h: New file.
14518         * config/nds32/nds32-elf.opt: New file.
14519         * config/nds32/nds32-linux.opt: New file.
14520         * config/nds32/nds32-fp-as-gp.c
14521         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
14522         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
14523         TARGET_LINUX_ABI.
14524         (nds32_asm_file_end): Ditto.
14525         (nds32_print_operand): Ditto.
14526         (nds32_insert_attributes): Ditto.
14527         (nds32_init_libfuncs): New function.
14528         (TARGET_HAVE_TLS): Define.
14529         (TARGET_INIT_LIBFUNCS): Define.
14530         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
14531         spec content.
14532         (TARGET_ELF): Apply different mcmodel setting.
14533         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
14534         been migrated into elf.h and linux.h files.
14535         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
14536         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
14537         (mcmodel): The content has been migrated into nds32-elf.opt and
14538         nds32-linux.opt files.
14539         * config/nds32/t-elf: New file.
14540         * config/nds32/t-linux: New file.
14542 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
14543             Shiva Chen  <shiva0217@gmail.com>
14545         * config/nds32/constants.md (unspec_volatile_element): Add
14546         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
14547         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
14548         optimization.
14549         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
14550         (make_pass_nds32_fp_as_gp): Declare.
14551         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
14552         optmization pass.
14553         (nds32_asm_function_end_prologue): Remove unused asm output.
14554         (nds32_asm_function_begin_epilogue): Remove unused asm output.
14555         (nds32_asm_file_start): Output necessary fp_as_gp information.
14556         (nds32_option_override): Adjust register usage.
14557         (nds32_expand_prologue): Consider fp_as_gp situation.
14558         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
14559         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
14560         (epilogue): Ditto.
14561         (return): Ditto.
14562         (simple_return): Ditto.
14563         (omit_fp_begin): Output special directive for fp_as_gp.
14564         (omit_fp_end): Output special directive for fp_as_gp.
14565         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
14566         mforbid-fp-as-gp): New options.
14568 2018-06-01  Mark Wielaard  <mark@klomp.org>
14570         * dwarf2out.c (dwarf2out_finish): Remove generation of
14571         DW_AT_loclists_base.
14573 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
14575         * gimple-ssa-store-merging.c: Include gimple-fold.h.
14576         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
14577         (struct merged_store_group): Add bit_insertion field.
14578         (dump_char_array): Use standard hexadecimal format.
14579         (merged_store_group::merged_store_group): Set bit_insertion to false.
14580         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
14581         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
14582         also print the mask in the dump file.
14583         (pass_store_merging::gate): Minor tweak.
14584         (imm_store_chain_info::coalesce_immediate): Fix wrong association
14585         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
14586         stores with INTEGER_CST stores.
14587         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
14588         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
14589         and use it throughout.  Generate bit insertion sequences if need be.
14590         (pass_store_merging::process_store): Remove redundant condition.
14591         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
14593 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
14595         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
14596         the 128-bit floating point types.  Fix function comment.
14598 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14600         * config/aarch64/aarch64-simd.md
14601         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
14602         mnemonics.
14603         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
14604         mnemonics.
14606 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
14608         PR tree-optimization/85989
14609         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
14610         variable.
14611         (backprop::intersect_uses): Check it when deciding whether this
14612         is a backedge reference.
14613         (backprop::process_block): Add each phi to m_visited_phis
14614         after visiting it, then clear it at the end.
14616 2018-06-01  Richard Biener  <rguenther@suse.de>
14618         * tree-vectorizer.h (vect_dr_stmt): New function.
14619         (vect_get_load_cost): Adjust.
14620         (vect_get_store_cost): Likewise.
14621         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
14622         Use vect_dr_stmt instead of DR_SMTT.
14623         (vect_record_base_alignments): Likewise.
14624         (vect_calculate_target_alignment): Likewise.
14625         (vect_compute_data_ref_alignment): Likewise and make static.
14626         (vect_update_misalignment_for_peel): Likewise.
14627         (vect_verify_datarefs_alignment): Likewise.
14628         (vector_alignment_reachable_p): Likewise.
14629         (vect_get_data_access_cost): Likewise.  Pass down
14630         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
14631         (vect_get_peeling_costs_all_drs): Likewise.
14632         (vect_peeling_hash_get_lowest_cost): Likewise.
14633         (vect_enhance_data_refs_alignment): Likewise.
14634         (vect_find_same_alignment_drs): Likewise.
14635         (vect_analyze_data_refs_alignment): Likewise.
14636         (vect_analyze_group_access_1): Likewise.
14637         (vect_analyze_group_access): Likewise.
14638         (vect_analyze_data_ref_access): Likewise.
14639         (vect_analyze_data_ref_accesses): Likewise.
14640         (vect_vfa_segment_size): Likewise.
14641         (vect_small_gap_p): Likewise.
14642         (vectorizable_with_step_bound_p): Likewise.
14643         (vect_prune_runtime_alias_test_list): Likewise.
14644         (vect_analyze_data_refs): Likewise.
14645         (vect_supportable_dr_alignment): Likewise.
14646         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
14647         (vect_gen_prolog_loop_niters): Likewise.
14648         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14649         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
14650         modify DR_STMT.
14651         (vect_recog_mask_conversion_pattern): Likewise.
14652         (vect_try_gather_scatter_pattern): Likewise.
14653         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
14654         to vect_get_store_cost.
14655         (vect_get_store_cost): Get stmt_info instead of DR.
14656         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
14657         (vect_get_load_cost): Get stmt_info instead of DR.
14659 2018-06-01  Richard Biener  <rguenther@suse.de>
14661         PR middle-end/86017
14662         * gimple-fold.c (var_decl_component_p): Also allow offsetted
14663         vars wrapped in MEM_REFs.
14665 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
14667         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
14668         Fix subreg tests so that we only return a choice between
14669         GENERAL_REGS and FP_REGS if the original classes included both.
14671 2018-06-01  Richard Biener  <rguenther@suse.de>
14673         PR ipa/85960
14674         * tree-ssa-structalias.c (get_function_part_constraint):
14675         Handle NULL fi->decl.
14676         (find_func_aliases_for_call): Properly handle indirect
14677         fi from direct call.
14678         (find_func_clobbers): Likewise.
14679         (ipa_pta_execute): Likewise.
14680         (create_variable_info_for): For functions that are ifunc_resolver
14681         resolve to a varinfo that contains the result of the resolver call.
14682         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
14683         aliases.
14685 2018-05-31  Michael Collison  <michael.collison@arm.com>
14687         * config/aarch64/aarch64.md:
14688         (*fix_to_zero_extenddfdi2): New pattern.
14689         * gcc.target/aarch64/fix_extend1.c: New testcase.
14691 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
14693         PR middle-end/78809
14694         PR middle-end/83026
14695         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
14696         and BUILT_IN_STRNCMP_EQ.
14697         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
14698         BUILT_IN_STRNCMP_EQ.
14699         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
14700         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
14701         (gimple_fold_builtin): Likewise.
14702         * tree-ssa-strlen.c (compute_string_length): New function.
14703         (determine_min_obsize): New function.
14704         (handle_builtin_string_cmp): New function to handle calls to
14705         string compare functions.
14706         (strlen_optimize_stmt): Add handling to builtin string compare
14707         calls.
14708         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
14709         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
14710         * tree.c (build_common_builtin_nodes): Add new defines of
14711         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
14713 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
14715         PR target/85984
14716         * bb-reorder.c (pass_partition_blocks::gate): Return false for
14717         functions with naked attribute.
14719 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
14721         * config/i386/sse.md (avx_vec_concat<mode>):
14722         Substitute concat_tg_mode mode attribute with xtg_mode.
14723         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
14724         (concat_tg_mode): Remove mode attribute.
14726 2018-05-31  Martin Sebor  <msebor@redhat.com>
14728         PR c/82063
14729         * calls.c (alloc_max_size): Correct a logic error/typo.
14730         Treat excessive arguments as infinite.  Warn for invalid arguments.
14731         * doc/invoke.texi (-Walloc-size-larger-than): Update.
14733 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
14735         PR target/85829
14736         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
14737         and movx for Haswell.
14739 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
14740             Cesar Philippidis  <cesar@codesourcery.com>
14742         PR middle-end/85879
14743         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
14744         when emitting error on private/firstprivate reductions.
14745         * omp-low.c (lower_omp_target): Avoid reference-type processing
14746         on pointers for firstprivate clause.
14748 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
14750         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
14751         (st1x2): Likewise.
14752         (st1x3): Likewise.
14753         * config/aarch64/aarch64-simd.md
14754         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
14755         (aarch64_ld1_x3_<mode>): Likewise
14756         (aarch64_st1x2<VALLDIF:mode>): Likewise
14757         (aarch64_st1_x2_<mode>): Likewise
14758         (aarch64_st1x3<VALLDIF:mode>): Likewise
14759         (aarch64_st1_x3_<mode>): Likewise
14760         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
14761         (vld1_s8_x3): Likewise.
14762         (vld1_u16_x3): Likewise.
14763         (vld1_s16_x3): Likewise.
14764         (vld1_u32_x3): Likewise.
14765         (vld1_s32_x3): Likewise.
14766         (vld1_u64_x3): Likewise.
14767         (vld1_s64_x3): Likewise.
14768         (vld1_f16_x3): Likewise.
14769         (vld1_f32_x3): Likewise.
14770         (vld1_f64_x3): Likewise.
14771         (vld1_p8_x3): Likewise.
14772         (vld1_p16_x3): Likewise.
14773         (vld1_p64_x3): Likewise.
14774         (vld1q_u8_x3): Likewise.
14775         (vld1q_s8_x3): Likewise.
14776         (vld1q_u16_x3): Likewise.
14777         (vld1q_s16_x3): Likewise.
14778         (vld1q_u32_x3): Likewise.
14779         (vld1q_s32_x3): Likewise.
14780         (vld1q_u64_x3): Likewise.
14781         (vld1q_s64_x3): Likewise.
14782         (vld1q_f16_x3): Likewise.
14783         (vld1q_f32_x3): Likewise.
14784         (vld1q_f64_x3): Likewise.
14785         (vld1q_p8_x3): Likewise.
14786         (vld1q_p16_x3): Likewise.
14787         (vld1q_p64_x3): Likewise.
14788         (vst1_s64_x2): Likewise.
14789         (vst1_u64_x2): Likewise.
14790         (vst1_f64_x2): Likewise.
14791         (vst1_s8_x2): Likewise.
14792         (vst1_p8_x2): Likewise.
14793         (vst1_s16_x2): Likewise.
14794         (vst1_p16_x2): Likewise.
14795         (vst1_s32_x2): Likewise.
14796         (vst1_u8_x2): Likewise.
14797         (vst1_u16_x2): Likewise.
14798         (vst1_u32_x2): Likewise.
14799         (vst1_f16_x2): Likewise.
14800         (vst1_f32_x2): Likewise.
14801         (vst1_p64_x2): Likewise.
14802         (vst1q_s8_x2): Likewise.
14803         (vst1q_p8_x2): Likewise.
14804         (vst1q_s16_x2): Likewise.
14805         (vst1q_p16_x2): Likewise.
14806         (vst1q_s32_x2): Likewise.
14807         (vst1q_s64_x2): Likewise.
14808         (vst1q_u8_x2): Likewise.
14809         (vst1q_u16_x2): Likewise.
14810         (vst1q_u32_x2): Likewise.
14811         (vst1q_u64_x2): Likewise.
14812         (vst1q_f16_x2): Likewise.
14813         (vst1q_f32_x2): Likewise.
14814         (vst1q_f64_x2): Likewise.
14815         (vst1q_p64_x2): Likewise.
14816         (vst1_s64_x3): Likewise.
14817         (vst1_u64_x3): Likewise.
14818         (vst1_f64_x3): Likewise.
14819         (vst1_s8_x3): Likewise.
14820         (vst1_p8_x3): Likewise.
14821         (vst1_s16_x3): Likewise.
14822         (vst1_p16_x3): Likewise.
14823         (vst1_s32_x3): Likewise.
14824         (vst1_u8_x3): Likewise.
14825         (vst1_u16_x3): Likewise.
14826         (vst1_u32_x3): Likewise.
14827         (vst1_f16_x3): Likewise.
14828         (vst1_f32_x3): Likewise.
14829         (vst1_p64_x3): Likewise.
14830         (vst1q_s8_x3): Likewise.
14831         (vst1q_p8_x3): Likewise.
14832         (vst1q_s16_x3): Likewise.
14833         (vst1q_p16_x3): Likewise.
14834         (vst1q_s32_x3): Likewise.
14835         (vst1q_s64_x3): Likewise.
14836         (vst1q_u8_x3): Likewise.
14837         (vst1q_u16_x3): Likewise.
14838         (vst1q_u32_x3): Likewise.
14839         (vst1q_u64_x3): Likewise.
14840         (vst1q_f16_x3): Likewise.
14841         (vst1q_f32_x3): Likewise.
14842         (vst1q_f64_x3): Likewise.
14843         (vst1q_p64_x3): Likewise.
14845 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
14847         * config/msp430/msp430.c (msp430_output_labelref): Prepend
14848         user_label_prefix to name.
14850         * tree-core.h: Update comment about the format of NAME string
14851         passed to handler in attribute_spec.
14853         * config/msp430/msp430.md: Remove erroneous subreg expression from
14854         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
14855         zero_extend{q,h}isi2.
14857 2018-05-30  Borislav Petkov  <bp@suse.de>
14859         * doc/extend.texi: Document some architecture specific
14860         constraints and sort entries.
14862 2018-05-30  Martin Sebor  <msebor@redhat.com>
14864         PR middle-end/85369
14865         * builtins.c (expand_builtin_stpcpy_1): New function.
14866         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
14867         only if the former succeeds.
14869 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
14871         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
14872         in saphira.
14874 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14876         * doc/invoke.texi (-flinker-output): Document
14878 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14880         * passes.c (ipa_write_summaries): Only modify statements if body
14881         is in memory.
14882         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
14883         incrementally linking.
14884         (ipa_passes): Likewise.
14885         * lto-cgraph.c (lto_output_node): When incrementally linking do not
14886         pass down resolution info.
14887         * common.opt (flag_incremental_link): Update info.
14888         * gcc.c (plugin specs): Turn flinker-output=* to
14889         -plugin-opt=-linker-output-known
14890         * toplev.c (compile_file): Also cut compilation when doing incremental
14891         link.
14892         * flag-types. (enum lto_partition_model): Add
14893         LTO_LINKER_OUTPUT_NOLTOREL.
14894         (invoke.texi): Add -flinker-output docs.
14895         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
14896         link same way as WPA; do not stream in dead initializers.
14898         * dwarf2out.c (dwarf2out_die_ref_for_decl,
14899         darf2out_register_external_decl): Support incremental link.
14901 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14903         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
14905 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
14907         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
14908         it down to simple_object_copy_lto_debug_sections.
14909         (run_gcc): Determine incremental LTO link time and configure
14910         lto1 into non-wpa mode, disable renaming of debug sections.
14912 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14914         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
14915         descriptions of various incorrectly documented functions.
14917 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
14919         Revert:
14920         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14921         address check not strict.
14923 2018-05-30  Richard Biener  <rguenther@suse.de>
14925         PR tree-optimization/85964
14926         * tracer.c (better_p): Drop initialized count check, we only
14927         call the function with initialized counts now.
14928         (find_best_successor): Do find a best edge if one
14929         has uninitialized count.
14930         (find_best_predecessor): Likewise.  Do BB frequency check only
14931         if count is initialized.
14933 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
14935         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
14936         (aarch64_ldrstr_offset_compare): New.
14937         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
14938         load/store orderings.
14939         (aarch64_gen_adjusted_ldpstp): Likewise.
14941 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
14943         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
14944         Check for subset of GENERAL_REGS and FP_REGS.
14945         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
14946         r=w alternative.
14948 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
14950         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
14951         and wi::to_poly_offset.  Add the current offset and then check
14952         whether the sum fits, rather than using an unchecked addition of
14953         a checked term.  Check for a shwi rather than a uhwi.
14954         * expr.c (get_bit_range): Use tree_to_poly_uint64.
14955         (store_constructor): Use poly_int_tree_p.
14956         (expand_expr_real_1): Likewise.
14957         * function.c (assign_temp): Likewise.
14958         * fold-const.c (const_binop): Use poly_int_tree_p and
14959         wi::to_poly_offset.
14960         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
14961         division.
14962         * ipa-icf-gimple.c (func_checker::compare_operand): Use
14963         to_poly_offset for MEM offsets.
14964         * ipa-icf.c (sem_variable::equals): Likewise.
14965         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
14966         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
14967         wi::to_poly_offset for BIT_FIELD_REF offsets.
14968         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
14969         wi::to_poly_offset.
14970         * var-tracking.c (emit_note_insn_var_location): Use
14971         tree_to_poly_uint64.
14973 2018-05-29  Jim Wilson  <jimw@sifive.com>
14975         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
14977 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
14979         PR target/85950
14980         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
14981         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
14982         sequence.
14983         (sse4_1_round<mode>2): Use nonimmediate_operand
14984         for operand 1 predicate.
14986 2018-05-29  Martin Sebor  <msebor@redhat.com>
14987             Richard Biener  <rguenther@suse.de>
14989         PR testsuite/85888
14990         * calls.c (get_size_range): Call determine_value_range instead
14991         of get_value_range..
14992         * tree-vrp.h (determine_value_range): Declared new function.
14993         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
14995 2018-05-29  Richard Biener  <rguenther@suse.de>
14997         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
14998         sure to use non-pattern stmts for get_earlier_stmt arguments.
14999         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
15000         called on pattern stmts.
15001         (get_later_stmt): Likewise.
15003 2018-05-29  Martin Liska  <mliska@suse.cz>
15005         PR gcov-profile/85759
15006         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
15007         env variables.
15009 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
15011         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
15012         VEC_UNPACK_*_EXPR.
15013         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
15014         VEC_PACK_*_EXPR.
15016         PR target/85918
15017         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15018         VEC_PACK_FLOAT_EXPR): New tree codes.
15019         * tree-pretty-print.c (op_code_prio): Handle
15020         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15021         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15022         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15023         * tree-inline.c (estimate_operator_cost): Likewise.
15024         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
15025         * fold-const.c (const_binop): Likewise.
15026         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
15027         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
15028         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15029         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
15030         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
15031         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
15032         * expr.c (expand_expr_real_2): Likewise.
15033         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
15034         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
15035         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
15036         optabs.
15037         * optabs.c (expand_widen_pattern_expr): For
15038         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
15039         sign from result type rather than operand's type.
15040         (expand_binop_directly): For vec_packu_float_optab and
15041         vec_packs_float_optab allow result type to be different from operand's
15042         type.
15043         * optabs-tree.c (optab_for_tree_code): Handle
15044         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15045         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
15046         * tree-vect-generic.c (expand_vector_operations_1):  Handle
15047         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
15048         VEC_PACK_FLOAT_EXPR.
15049         * tree-vect-stmts.c (supportable_widening_operation): Handle
15050         FIX_TRUNC_EXPR.
15051         (supportable_narrowing_operation): Handle FLOAT_EXPR.
15052         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
15053         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
15054         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
15055         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
15056         mode attributes.
15057         (vec_pack<floatprefix>_float_<mode>): New expander.
15058         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
15059         attributes.
15060         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
15061         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
15062         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
15063         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
15064         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
15065         Document.
15066         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
15067         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
15068         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
15069         VEC_PACK_FLOAT_EXPR): Document.
15071 2018-05-29  Richard Biener  <rguenther@suse.de>
15073         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
15074         member.
15075         (stmt_vec_info_vec): Make pointer.
15076         (init_stmt_vec_info_vec): Remove.
15077         (free_stmt_vec_info_vec): Likewise.
15078         (set_stmt_vec_info_vec): New function.
15079         (free_stmt_vec_infos): Likewise.
15080         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
15081         (set_vinfo_for_stmt): Likewise.
15082         (get_earlier_stmt): Likewise.
15083         (get_later_stmt): Likewise.
15084         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
15085         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
15086         (vec_info::~vec_info): Free stmt_vec_infos.
15087         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
15088         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
15089         (pass_slp_vectorize::execute): Likewise.
15090         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
15091         (free_stmt_vec_info_vec): Likewise.
15092         (set_stmt_vec_info_vec): New function.
15093         (free_stmt_vec_infos): Likewise.
15094         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
15095         the global stmt_vec_info_vec.
15096         * tree-parloops.c (gather_scalar_reductions): Use
15097         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
15098         vector.
15100 2018-05-29  Richard Biener  <rguenther@suse.de>
15102         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
15104 2018-05-29  Martin Liska  <mliska@suse.cz>
15105             David Malcolm  <dmalcolm@redhat.com>
15107         * vec.c (test_reverse): New.
15108         (vec_c_tests): Add new test.
15109         * vec.h (vl_ptr>::reverse): New function.
15111 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
15113         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
15115         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
15116         and later.
15118 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15120         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
15122 2018-05-28  Richard Biener  <rguenther@suse.de>
15124         PR tree-optimization/85933
15125         * tree-vect-data-refs.c (vect_record_base_alignments): Only
15126         look at stmts marked as vectorizable.
15128 2018-05-28  Richard Biener  <rguenther@suse.de>
15130         PR tree-optimization/85934
15131         * tree-vect-generic.c (expand_vector_operations_1): Hoist
15132         vector boolean check before scalar optimization.
15134 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
15136         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
15137         for armv5te.
15139 2018-05-28  Mark Wielaard  <mark@klomp.org>
15141         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
15142         if it is an expression containing a minus sign.
15144 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
15146         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
15148 2018-05-27  Paul Koning  <ni1d@arrl.net>
15150         * config/pdp11/pdp11.md (truncsihi2): Remove.
15152 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
15153             Chung-Ju Wu  <jasonwucj@gmail.com>
15155         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
15156         implementation.
15157         (unaligned_store_dw): Ditto.
15158         * config/nds32/nds32-memory-manipulation.c
15159         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
15160         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
15161         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
15162         (emit_setmem_word_loop): Rename to ...
15163         (emit_setmem_doubleword_loop): ... this.
15164         (nds32_gen_dup_4_byte_to_word_value): New function.
15165         (nds32_gen_dup_8_byte_to_double_word_value): New function.
15166         (nds32_expand_setmem_loop): Refine implementation.
15167         (nds32_expand_setmem_loop_v3m): Ditto.
15168         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
15169         pattern.
15171 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
15173         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
15175 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
15177         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
15178         (nds32_init_machine_status): Initialize machine->attr_naked_p and
15179         machine->attr_no_prologue_p.
15180         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
15181         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
15182         (nds32_expand_epilogue): Consider attr_naked_p.
15183         (nds32_expand_epilogue_v3pop): Likewise.
15184         (nds32_can_use_return_insn): Likewise.
15185         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
15186         attr_no_prologue_p fields.
15187         * config/nds32/nds32.opt (mret-in-naked-func): New option.
15189 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
15191         PR target/85918
15192         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
15193         attributes.
15194         * config/i386/sse.md
15195         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
15196         Rename to ...
15197         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
15198         ... this.
15199         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
15200         Rename to ...
15201         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
15202         ... this.
15203         (*<floatsuffix>floatv2div2sf2): Rename to ...
15204         (*float<floatunssuffix>v2div2sf2): ... this.
15205         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
15206         (float<floatunssuffix>v2div2sf2_mask): ... this.
15207         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
15208         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
15209         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
15210         to ...
15211         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
15212         ... this.
15213         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
15214         Rename to ...
15215         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
15216         ... this.
15217         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
15218         Rename to ...
15219         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
15220         ... this.
15221         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
15222         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
15223         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
15224         gen_ufix_truncv8dfv8si2.
15225         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
15226         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
15227         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
15228         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
15229         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
15230         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
15231         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
15232         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
15234 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
15236         PR target/85900
15237         PR target/85345
15238         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
15240 2018-05-25  Jim Wilson  <jimw@sifive.com>
15242         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
15243         * config/riscv/riscv.c (struct machine_function): Add
15244         interrupt_handler_p and attribute_checked_p fields.
15245         (riscv_attribute_table): Add interrupt.
15246         (riscv_interrupt_type_p): New.
15247         (riscv_save_reg_p): Save extra regs for interrupt handler.
15248         (riscv_use_save_libcall): Return false  for interrupt handler.
15249         (riscv_first_stack_step): Add forward declaration.
15250         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
15251         for interrupt handler with large frame.  Use it for saved reg list.
15252         (riscv_expand_prologue): Move flag_stack_usage_info support to
15253         eliminate duplication.
15254         (riscv_expand_epilogue): Generate mret for interrupt handler.
15255         (riscv_epilogue_uses): New.
15256         (riscv_can_use_return_insn): Return false for interrupt handler.
15257         (riscv_function_ok_for_sibcall): Likewise.
15258         (riscv_set_current_function): Add interrupt handler support.
15259         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
15260         * config/riscv/riscv.md (UNSPECV_MRET): New.
15261         (GP_REGNUM): New.
15262         (riscv_frflags, riscv_fsflags): Use tab after opcode.
15263         (riscv_mret): New.
15264         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
15266 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
15268         PR tree-optimization/85712
15269         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
15270         this candidate has already been replaced in-situ by a copy.
15272 2018-05-25  Jason Merrill  <jason@redhat.com>
15274         PR c++/80485 - inline function non-zero address.
15275         * symtab.c (nonzero_address): Check DECL_COMDAT.
15277 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
15279         PR target/83628
15280         * config/alpha/alpha.md (ashlsi3): New insn pattern.
15281         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
15282         extension of SImode operation.  Use const123_operand predicate.
15283         (*saddsi_1): Remove.
15284         (*saddl_se_1): Ditto.
15285         (*ssubsi_1): Ditto.
15286         (*ssubl_se_1): Ditto.
15287         * config/alpha/predicates.md (const123_operand): New predicate.
15288         * config/alpha/constraints.md (P): Use IN_RANGE.
15290 2018-05-25  Richard Biener  <rguenther@suse.de>
15292         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
15293         defaulted to true.
15294         (ref_maybe_used_by_stmt_p): Likewise.
15295         (stmt_may_clobber_ref_p): Likewise.
15296         (stmt_may_clobber_ref_p_1): Likewise.
15297         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
15298         and pass it along.
15299         (ref_maybe_used_by_stmt_p): Likewise.
15300         (stmt_may_clobber_ref_p): Likewise.
15301         (stmt_may_clobber_ref_p_1): Likewise.
15302         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
15303         the alias oracle to disambiguate DRs with stmts DR analysis
15304         couldn't handle.
15305         (vect_analyze_data_refs): Do not give up on not analyzable
15306         DRs for BB vectorization.  Remove code truncating the dataref
15307         vector.
15309 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
15311         PR target/85832
15312         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
15313         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
15314         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
15316 2018-05-25  Richard Biener  <rguenther@suse.de>
15318         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
15319         function, combining stmt data ref gathering and fatal analysis
15320         parts.
15321         (vect_analyze_data_refs): Remove now redudnant code and simplify.
15322         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
15323         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
15324         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
15325         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
15327 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
15329         PR tree-optimization/85720
15330         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
15331         SCC if all partitions are builtins.
15332         (version_loop_by_alias_check): New parameter.  Generate cancelable
15333         runtime alias check if all partitions are builtins.
15334         (distribute_loop): Update call to above function.
15336 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
15338         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
15339         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
15340         (parm_default_def_partition_arg): Ditto.
15341         (set_parm_default_def_partition): Ditto.
15342         (get_parm_default_def_partitions): Ditto and make it static.
15343         (get_undefined_value_partitions): Ditto and make it static.
15344         (remove_ssa_form): Refactor call to init_var_map here.
15345         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
15346         computation for loop region.
15347         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
15348         (register_default_def): Delete.
15349         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
15350         (parm_default_def_partition_arg): Ditto.
15351         (set_parm_default_def_partition): Ditto.
15352         (get_parm_default_def_partitions): Ditto and make it static.
15353         (get_undefined_value_partitions): Ditto and make it static.
15354         (coalesce_with_default, coalesce_with_default): Update comment.
15355         (create_coalesce_list_for_region): New func factored out from
15356         create_outofssa_var_map.
15357         (populate_coalesce_list_for_outofssa): New func factored out from
15358         create_outofssa_var_map and coalesce_ssa_name.
15359         (create_outofssa_var_map): Delete.
15360         (coalesce_ssa_name): Refactor to support live range computation.
15361         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
15362         (get_parm_default_def_partitions): Delete.
15363         (get_undefined_value_partitions): Ditto.
15364         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
15365         computation for loop region.
15366         (new_tree_live_info, loe_visit_block): Ditto.
15367         (live_worklist, set_var_live_on_entry): Ditto.
15368         (calculate_live_on_exit, verify_live_on_entry): Ditto.
15369         * tree-ssa-live.h (struct _var_map): New fields.
15370         (init_var_map): Change decl.
15371         (region_contains_p): New.
15373 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
15375         * tree-ssa-live.h (live_merge_and_clear): Delete.
15377 2018-05-25  Richard Biener  <rguenther@suse.de>
15379         PR c++/85912
15380         * tree-dump.c (dequeue_and_dump): Remove access to removed
15381         operand 2 of a SWITCH_EXPR.
15383 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15385         * doc/sourcebuild.texi (vect_double_cond_arith): Include
15386         multiplication and division.
15387         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
15388         (cond_udiv@var{m}, cond_umod@var{m}): Document.
15389         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
15390         (cond_udiv_optab, cond_umod_optab): New optabs.
15391         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
15392         (IFN_COND_RDIV): New internal functions.
15393         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
15394         TRUNC_MOD_EXPR and RDIV_EXPR.
15395         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
15396         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
15397         New unspecs.
15398         (SVE_INT_BINARY): Include mult.
15399         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
15400         (optab, sve_int_op): Handle mult.
15401         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
15402         UNSPEC_COND_DIV.
15403         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
15404         for SVE_INT_BINARY_SD.
15406 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15408         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
15409         (optab, sve_int_op): Handle div and udiv.
15410         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
15411         for SVE_INT_BINARY_SD.
15412         (*<optab><mode>3): New insn for the same.
15414 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15416         * tree-vect-patterns.c: Include predict.h.
15417         (vect_recog_divmod_pattern): Restrict check for division support
15418         to when optimizing for size.
15420 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15422         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
15423         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
15424         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
15425         (gimple_match_op::set_op): Likewise.
15426         (gimple_resimplify4): Declare.
15427         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
15428         (expr::gen_transform): Likewise.
15429         (decision_tree::gen): Generate a simplification routine for 4 operands.
15430         * gimple-match-head.c (gimple_simplify): Add an overload for
15431         4 operands.  In the top-level function, handle up to 4 call
15432         arguments and call gimple_resimplify4.
15433         (gimple_resimplify4): New function.
15434         (build_call_internal): Pass a fourth operand.
15435         (maybe_push_to_seq): Likewise.
15436         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
15437         Fold VEC_COND_EXPRs of an operation and a default value into
15438         an IFN_COND_* function if possible.
15439         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
15440         New unspecs.
15441         (SVE_COND_FP_BINARY): Include them.
15442         (optab, sve_fp_op): Handle them.
15443         (SVE_INT_BINARY_REV): New code iterator.
15444         (SVE_COND_FP_BINARY_REV): New int iterator.
15445         (commutative): New int attribute.
15446         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
15447         Declare.
15448         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
15449         function.
15450         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
15451         (*cond_<optab><mode>): New patterns for reversed operands.
15453 2018-05-25  Richard Biener  <rguenther@suse.de>
15455         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
15456         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
15457         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
15458         (STMT_VINFO_GROUPED_ACCESS): Adjust.
15459         * tree-vect-data-refs.c (everywhere): Adjust users.
15460         * tree-vect-loop.c (everywhere): Likewise.
15461         * tree-vect-slp.c (everywhere): Likewise.
15462         * tree-vect-stmts.c (everywhere): Likewise.
15463         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
15465 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15467         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
15468         Rename to...
15469         (gcc_cv_as_section_exclude): ... this.
15470         Try Solaris as #exclude syntax.
15471         * configure: Regenerate.
15472         * config.in: Regenerate.
15473         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
15474         SECTION_EXCLUDE.
15475         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
15476         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
15478         * varasm.c (default_elf_asm_named_section): Don't check if
15479         HAVE_GAS_SECTION_EXCLUDE is defined.
15481 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15483         * doc/md.texi: Update the documentation of the cond_* optabs
15484         to mention the new final operand.  Fix GET_MODE_NUNITS call.
15485         Describe the scalar case too.
15486         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
15487         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
15488         instead of 2.
15489         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
15490         (get_conditional_internal_fn): Update comment.
15491         * tree-vect-loop.c (vectorizable_reduction): Pass the original
15492         accumulator value as a final argument to conditional functions.
15493         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
15494         a define_expand and add an "else" operand.  Assert for now that
15495         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
15496         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
15497         (*cond_<optab><mode>): New patterns.
15498         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
15499         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
15500         (UNSPEC_COND_EOR): Delete.
15501         (optab): Remove associated mappings.
15502         (SVE_INT_BINARY): New code iterator.
15503         (sve_int_op): Remove int attribute and add "minus" to the code
15504         attribute.
15505         (SVE_COND_INT_OP): Delete.
15506         (SVE_COND_FP_OP): Rename to...
15507         (SVE_COND_FP_BINARY): ...this.
15509 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
15511         * optabs.c (can_reuse_operands_p): New function.
15512         (maybe_legitimize_operands): Try to reuse the results for
15513         earlier operands.
15515 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
15517         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
15518         Add {q} suffix to insn mnemonic.
15520 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
15522         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
15523         (msp430_warn_func_return): New.
15525 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
15527         * fold-const.c (tree_nonzero_bits): New function.
15528         * fold-const.h (tree_nonzero_bits): Likewise.
15529         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
15530         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
15532 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
15534         PR target/85900
15535         PR target/85345
15536         * varasm.c (assemble_alias): Check ifunc_resolver only on
15537         FUNCTION_DECL.
15539 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
15541         PR target/85903
15542         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
15543         when memory input operand is handled.
15545 2018-05-24  Luis Machado  <luis.machado@linaro.org>
15547         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
15548         global.
15549         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
15551 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
15553         * match.pd: Delay FMA folds until after vectorization.
15555 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15557         PR target/83009
15558         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
15559         address check not strict.
15561 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
15563         * gimple-match.h (gimple_match_op): New class.
15564         (mprts_hook): Replace parameters with a gimple_match_op *.
15565         (maybe_build_generic_op): Likewise.
15566         (gimple_simplified_result_is_gimple_val): Replace parameters with
15567         a const gimple_match_op *.
15568         (gimple_simplify): Replace code_helper * and tree * parameters with
15569         a gimple_match_op * parameter.
15570         (gimple_resimplify1): Replace code_helper *, tree and tree *
15571         parameters with a gimple_match_op * parameter.
15572         (gimple_resimplify2): Likewise.
15573         (gimple_resimplify3): Likewise.
15574         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
15575         parameters with a gimple_match_op * parameter.
15576         * gimple-match-head.c (gimple_simplify): Change prototypes of
15577         auto-generated functions to take a gimple_match_op * instead of
15578         separate code_helper * and tree * parameters.  Make the same
15579         change in the top-level overload and update calls to the
15580         gimple_resimplify routines.  Update calls to the auto-generated
15581         functions and to maybe_push_res_to_seq in the publicly-facing
15582         operation-specific gimple_simplify overloads.
15583         (gimple_match_op::MAX_NUM_OPS): Define.
15584         (gimple_resimplify1): Replace rcode and ops with a single res_op
15585         parameter.  Update call to gimple_simplify.
15586         (gimple_resimplify2): Likewise.
15587         (gimple_resimplify3): Likewise.
15588         (mprts_hook): Replace parameters with a gimple_match_op *.
15589         (maybe_build_generic_op): Likewise.
15590         (build_call_internal): Replace type, nargs and ops with
15591         a gimple_match_op *.
15592         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
15593         with a single gimple_match_op *.  Update calls to mprts_hook,
15594         build_call_internal and gimple_simplified_result_is_gimple_val.
15595         Factor out code that is common to the tree_code and combined_fn cases.
15596         * genmatch.c (expr::gen_transform): Replace tem_code and
15597         tem_ops with a gimple_match_op called tem_op.  Update calls
15598         to the gimple_resimplify functions and maybe_push_res_to_seq.
15599         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
15600         res_ops.  Update call to the gimple_resimplify functions.
15601         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
15602         (decision_tree::gen): Make the functions take a gimple_match_op *
15603         called res_op instead of separate res_code and res_ops parameters.
15604         Update call accordingly.
15605         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
15606         and ops with a single res_op parameter.  Update calls to
15607         maybe_build_generic_op and maybe_push_res_to_seq.
15608         (fold_stmt_1): Update calls to gimple_simplify and
15609         replace_stmt_with_simplification.
15610         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
15611         and gimple_simplified_result_is_gimple_val.
15612         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
15613         gimple_simplify.
15614         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
15615         with a gimple_match_op *.
15616         (vn_nary_build_or_lookup): Likewise.  Update call to
15617         vn_nary_build_or_lookup_1.
15618         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
15619         gimple_match_op *.  Update calls to the gimple_resimplify routines
15620         and to gimple_simplified_result_is_gimple_val.
15621         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
15622         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
15623         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
15624         (visit_nary_op): Likewise.
15625         (visit_reference_op_load): Likewise.
15627 2018-05-23  Luis Machado  <luis.machado@linaro.org>
15629         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
15630         modifier for printing the step amount.
15632 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
15634         PR target/78849
15635         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
15636         types.
15638 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
15640         * doc/sourcebuild.texi (Endianness): New subsubsection.
15642 2018-05-23  Luis Machado  <luis.machado@linaro.org>
15644         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15645         <prefetch_dynamic_strides>: New const bool field.
15646         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15647         prefetch_dynamic_strides.
15648         (exynosm1_prefetch_tune): Likewise.
15649         (thunderxt88_prefetch_tune): Likewise.
15650         (thunderx_prefetch_tune): Likewise.
15651         (thunderx2t99_prefetch_tune): Likewise.
15652         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
15653         false.
15654         (aarch64_override_options_internal): Update to set
15655         PARAM_PREFETCH_DYNAMIC_STRIDES.
15656         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15657         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15658         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15659         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15660         prefetch-dynamic-strides setting.
15662 2018-05-23  Luis Machado  <luis.machado@linaro.org>
15664         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15665         <minimum_stride>: New const int field.
15666         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15667         minimum_stride field defaulting to -1.
15668         (exynosm1_prefetch_tune): Likewise.
15669         (thunderxt88_prefetch_tune): Likewise.
15670         (thunderx_prefetch_tune): Likewise.
15671         (thunderx2t99_prefetch_tune): Likewise.
15672         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
15673         <default_opt_level>: Set to 3.
15674         (aarch64_override_options_internal): Update to set
15675         PARAM_PREFETCH_MINIMUM_STRIDE.
15676         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15677         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15678         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15679         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15680         stride is constant and is below the minimum stride threshold.
15682 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15684         * config/arm/arm-cpus.in (mode26): Delete.
15685         (armv4): Delete mode26 reference.
15686         * config/arm/arm.c (arm_configure_build_target): Delete use of
15687         isa_bit_mode26.
15689 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
15691         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
15692         New insn pattern.
15693         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
15694         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
15695         for non-SSE modes.
15696         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
15697         (floatunsdidf2): Ditto.
15699 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
15701         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
15702         (fixuns_trunc<mode>si2_avx512f): Ditto.
15703         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
15704         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
15705         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
15707 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
15709         PR rtl-optimization/79985
15710         * df-scan.c (df_insn_refs_collect): Remove special case for
15711         global registers and asm statements.
15713 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
15715         * extend.texi (Global Register Variables): Rewrite the bullet list.
15716         Note that the register is available for allocation. Note that access
15717         via inline asm must use constraints. Add note about async-signal
15718         handlers. Remove paragraph about automagic register selection.
15720 2018-05-23  Richard Biener  <rguenther@suse.de>
15722         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
15723         of fixed offset from memset VN.
15725 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
15727         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
15728         first_interp field.
15729         (alloc_cand_and_find_basis): Initialize first_interp field.
15730         (slsr_process_mul): Modify first_interp field.
15731         (slsr_process_add): Likewise.
15732         (slsr_process_cast): Modify first_interp field for each new
15733         interpretation.
15734         (slsr_process_copy): Likewise.
15735         (dump_candidate): Dump first_interp field.
15736         (replace_mult_candidate): Process all interpretations, not just
15737         subsequent ones.
15738         (replace_rhs_if_not_dup): Likewise.
15739         (replace_one_candidate): Likewise.
15741 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
15743         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
15744         Add new boolean.
15745         (aarch64_needs_frame_chain): New function.
15746         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
15748 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
15750         PR target/84882
15751         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
15752         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
15753         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
15754         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
15755         as true for strict-align.
15756         (aarch64_can_inline_p): Perform checks even when callee has no
15757         attributes to check for strict alignment.
15758         * doc/extend.texi (AArch64 Function Attributes): Document
15759         no-strict-align.
15760         * doc/invoke.texi: (AArch64 Options): Likewise.
15762 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
15764         PR tree-optimization/85853
15765         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
15766         the handling of the root of the node to...
15767         (vect_slp_analyze_node_operations_1): ...this new function,
15768         and run the whole thing with the child nodes' def types
15769         set according to their SLP node's def type.
15771 2018-05-23  Richard Biener  <rguenther@suse.de>
15773         PR middle-end/85874
15774         * tree-data-ref.c (create_runtime_alias_checks): Defer
15775         and ignore overflow warnings.
15777 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
15779         PR tree-optimization/85822
15780         * tree-vrp.c (is_masked_range_test): Fix handling of negative
15781         constants.
15783 2018-05-23  Richard Biener  <rguenther@suse.de>
15785         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
15786         memset constants via native_interpret_expr.
15788 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
15790         PR target/85345
15791         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
15792         attribute.
15793         (cgraph_node::create_alias): Likewise.
15794         (cgraph_node::get_availability): Check ifunc_resolver instead
15795         of looking up ifunc attribute.
15796         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
15797         * varasm.c (do_assemble_alias): Likewise.
15798         (assemble_alias): Likewise.
15799         (default_binds_local_p_3): Likewise.
15800         * cgraph.h (cgraph_node): Add ifunc_resolver.
15801         (cgraph_node::only_called_directly_or_aliased_p): Return false
15802         for IFUNC resolver.
15803         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
15804         attribute.
15805         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
15806         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
15807         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
15808         instead of looking up ifunc attribute.
15810 2018-05-22  Luis Machado  <luis.machado@linaro.org>
15812         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
15814 2018-05-22  Martin Sebor  <msebor@redhat.com>
15816         PR middle-end/85359
15817         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
15818         only when expasion succeeds.
15819         (expand_builtin_strcmp): Same.
15820         (expand_builtin_strncmp): Same.
15822 2018-05-22  Martin Sebor  <msebor@redhat.com>
15824         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
15826 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
15827             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15829         * config/aarch64/aarch64-ldpstp.md: Replace uses of
15830         aarch64_mem_pair_operand with memory_operand and delete operand swapping
15831         code.
15832         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
15833         Add check for legitimate_address.
15834         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
15835         (aarch64_swap_ldrstr_operands): New.
15836         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
15837         Define prototype.
15839 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
15840             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15842         * config/aarch64/aarch64.md: New patterns to generate stp
15843         and ldp.
15844         (store_pair_sw, store_pair_dw): New patterns to generate stp for
15845         single words and double words.
15846         (load_pair_sw, load_pair_dw): Likewise.
15847         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
15848         Delete.
15849         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
15850         Delete.
15851         * config/aarch64/aarch64-ldpstp.md: Modify peephole
15852         for different mode ldpstp and add peephole for merged zero stores.
15853         Likewise for loads.
15854         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
15855         Add size check.
15856         (aarch64_gen_store_pair): Rename calls to match new patterns.
15857         (aarch64_gen_load_pair): Rename calls to match new patterns.
15858         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
15859         (load_pair<DREG:mode><DREG2:mode>): ... This.
15860         (store_pair<mode>): Rename to...
15861         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
15862         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
15863         New mode iterators.
15864         (V_INT_EQUIV): Handle SImode.
15865         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
15866         New predicate.
15868 2018-05-22  Martin Sebor  <msebor@redhat.com>
15870         PR c/85623
15871         * calls.c (maybe_warn_nonstring_arg): Use string length to set
15872         or ajust the presumed bound on an operation to avoid unnecessary
15873         warnings.
15875 2018-05-22  Martin Sebor  <msebor@redhat.com>
15877         PR tree-optimization/85826
15878         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
15879         assuming that a DECL necesarily has a constant size.
15881 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
15883         PR middle-end/85862
15884         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
15886 2018-05-22  Richard Biener  <rguenther@suse.de>
15888         PR tree-optimization/85834
15889         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
15890         non-constant and non-zero memset arguments.
15892 2018-05-22  Martin Liska  <mliska@suse.cz>
15894         PR ipa/85607
15895         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
15897 2018-05-22  Richard Biener  <rguenther@suse.de>
15899         PR tree-optimization/85863
15900         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
15901         comparisons when vectype is specified.
15902         (vectorizable_condition): Do not specify vectype for
15903         vect_is_simple_cond when SLP vectorizing.
15905 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
15907         PR target/85657
15908         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
15909         define __ibm128 as long double.
15910         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
15911         as a distinct type when IEEE 128-bit support is enabled.
15912         (init_float128_ieee): Fix up conversions between IFmode and IEEE
15913         128-bit types to use the correct functions.
15914         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
15915         convert between 128-bit floating point types that have different
15916         modes but the same representation, instead of using gen_lowpart to
15917         makean alias.
15918         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
15919         KFmode.
15920         (IFKF_reg): New attributes to give the register constraints for
15921         IFmode and KFmode.
15922         (extend<mode>tf2_internal): New insns to mark an explicit
15923         conversion between 128-bit floating point types that have a
15924         different mode but share the same representation.
15926 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
15928         PR tree-optimization/85814
15929         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
15930         a null return from get_strinfo when unsharing the next
15931         strinfo in the chain.
15933 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
15935         PR gcc/84923
15936         * varasm.c (weak_finish): Clean up weak_decls.
15938 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15940         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
15941         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
15942         UNSPEC_UADALP values.
15943         * config/aarch64/iterators.md (ABAL): New int iterator.
15944         (ABDL2): Likewise.
15945         (ADALP): Likewise.
15946         (sur): Add mappings for the above.
15947         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
15948         New define_insn.
15949         (aarch64_<sur>abal<mode>_4): Likewise.
15950         (aarch64_<sur>adalp<mode>_3): Likewise.
15951         (<sur>sadv16qi): New define_expand.
15953 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
15955         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
15956         (*movdf_internal): Ditto.
15957         (*rcpsf2_sse): Ditto.
15958         (*rsqrtsf2_sse): Ditto.
15959         (*sqrt<mode>2_sse): Ditto.
15961 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
15963         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
15964         eor3q<mode>4.
15965         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
15966         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
15967         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
15968         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
15969         vbcaxq_s64): New.
15970         * config/aarch64/arm_neon.h: Likewise.
15971         * config/aarch64/iterators.md (VQ_I): New.
15973 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
15975         * config.gcc: Add arc/t-multilib-linux to tmake_file for
15976         arc*-*-linux*.
15977         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
15978         MULTILIB_DIRNAMES
15980 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
15982         * config/nds32/constraints.md (S): New constraint.
15983         * config/nds32/nds32.md (call_internal): Use constraint S.
15984         (call_value_internal): Likewise.
15985         (sibcall_internal): Likewise.
15986         (sibcall_value_internal): Likewise.
15988 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
15989             Chung-Ju Wu  <jasonwucj@gmail.com>
15991         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
15992         into consideration.
15994 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
15995             Chung-Ju Wu  <jasonwucj@gmail.com>
15997         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
15998         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
15999         (nds32_rtx_costs_impl): Simplify.
16000         (nds32_address_cost_impl): Simplify.
16001         (nds32_init_rtx_costs): New function.
16002         (nds32_rtx_costs_speed_prefer): Likewise.
16003         (nds32_rtx_costs_size_prefer): Likewise.
16004         (nds32_address_cost_speed_prefer): Likewise.
16005         (nds32_address_cost_speed_fwprop): Likewise.
16006         (nds32_address_cost_size_prefer): Likewise.
16007         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
16008         * config/nds32/nds32.c (nds32_option_override): Use
16009         nds32_init_rtx_costs function.
16011 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
16013         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
16014         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
16015         (TARGET_PIPELINE_N8): Likewise.
16016         (TARGET_PIPELINE_N10): Likewise.
16017         (TARGET_PIPELINE_N13): Likewise.
16018         (TARGET_PIPELINE_GRAYWOLF): Likewise.
16020 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
16022         * config/nds32/nds32-fpu.md: Update copyright year.
16024 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16026         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
16028 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16030         * config/nds32/nds32.c
16031         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
16032         * config/nds32/nds32.opt (minline-asm-r15): New option.
16034 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16036         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
16037         MASK_HW_ABS.
16038         * config/nds32/nds32.md (abssi2): New pattern.
16040 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
16042         * config/i386/i386.md (rex64namesuffix): New mode attribute.
16043         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
16044         Merge insn pattern from sse_cvtsi2ss<round_name> and
16045         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
16046         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
16047         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
16048         using SWI48 mode iterator.
16049         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
16050         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
16051         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
16052         pattern from sse_cvttss2si<round_saeonly_name>
16053         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
16054         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
16055         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
16056         using SWI48 mode iterator.
16057         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
16058         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
16059         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
16060         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
16061         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
16062         using SWI48 mode iterator.
16063         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
16064         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
16065         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
16066         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
16067         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
16068         SWI48 mode iterator.
16069         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
16070         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
16071         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
16072         pattern from sse_cvttsd2si<round_saeonly_name>
16073         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
16075 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16077         * config/nds32/nds32-md-auxiliary.c
16078         (nds32_valid_smw_lwm_base_p): Refine.
16079         (nds32_output_smw_single_word): Refine.
16080         (nds32_output_smw_double_word): New.
16081         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
16083 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16085         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
16086         (nds32_output_stack_pop): Refine.
16087         (nds32_expand_unaligned_load): Refine.
16088         (nds32_expand_unaligned_store): Refine.
16090 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
16091             Chung-Ju Wu  <jasonwucj@gmail.com>
16093         * config/nds32/constants.md: Add TP_REGNUM constant.
16094         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
16095         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
16096         UNSPEC_ADD32.
16097         * config/nds32/nds32-doubleword.md: Consider flag_pic.
16098         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
16099         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
16100         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
16101         and PIC code generation.
16102         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
16103         code generation.
16104         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
16105         optimization.
16106         * config/nds32/nds32.md: Support TLS and PIC.
16107         * config/nds32/nds32.c: Support TLS and PIC.
16108         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
16109         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
16110         predicate.
16112 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
16114         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
16115         mode with E_ prefix.
16117 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
16118             Chung-Ju Wu  <jasonwucj@gmail.com>
16120         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
16121         * config/nds32/nds32-md-auxiliary.c
16122         (symbolic_reference_mentioned_p): New.
16123         (nds32_legitimize_ict_address): New.
16124         (nds32_expand_ict_move): New.
16125         (nds32_indirect_call_referenced_p): New.
16126         (nds32_symbol_binds_local_p): Delete.
16127         (nds32_long_call_p): Modify.
16128         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
16129         * config/nds32/nds32-protos.h
16130         (symbolic_reference_mentioned_p): Declare.
16131         (nds32_legitimize_ict_address): Declare.
16132         (nds32_expand_ict_move): Declare.
16133         (nds32_indirect_call_referenced_p): Declare.
16134         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
16135         (nds32_relax_group): Use nds32_ict_const_p as condition.
16136         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
16137         (nds32_asm_file_start): Output ict_model directive in asm code.
16138         (nds32_legitimate_address_p): Consider indirect call.
16139         (nds32_print_operand): Consider indirect call.
16140         (nds32_print_operand_address): Consider indirect call.
16141         (nds32_insert_attributes): Handle "indirect_call" attribute.
16142         (TARGET_LEGITIMATE_ADDRESS_P): Define.
16143         (TARGET_LEGITIMATE_CONSTANT_P): Define.
16144         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
16145         (TARGET_DELEGITIMIZE_ADDRESS): Define.
16146         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
16147         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
16148         (TARGET_ICT_MODEL_SMALL): Define.
16149         (TARGET_ICT_MODEL_LARGE): Define.
16150         * config/nds32/nds32.md (movsi): Consider ict model.
16151         (call, call_value): Consider ict model.
16152         (sibcall, sibcall_value): Consider ict model.
16153         * config/nds32/nds32.opt (mict-model): New option.
16154         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
16155         model.
16157 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
16158             Monk Chiang  <sh.chiang04@gmail.com>
16159             Jim Wilson <jimw@sifive.com>
16161         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
16162         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
16163         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
16164         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
16165         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
16166         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
16167         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
16168         compute save_libcall_adjustment properly.
16169         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
16170         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
16171         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
16172         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
16173         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
16174         (ABI_SPEC): Handle mabi=ilp32e.
16175         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
16176         (RVE): Add RVE mask.
16177         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
16178         <-march>: Add rv32e as an example.
16180 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
16182         PR c++/82899
16183         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
16184         (intra_create_variable_infos): Handle C++ constructors.
16186 2018-05-18  Martin Liska  <mliska@suse.cz>
16188         * passes.def: Remove a redundant pass.
16190 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
16192         PR bootstrap/85838
16193         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
16195 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16197         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
16198         (ARMv4): Update.
16199         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
16200         (ARMv6m): Update.
16201         (armv2, armv2a, armv3, armv3m): Delete architectures.
16202         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
16203         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
16204         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
16205         Delete cpus.
16206         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
16207         (*mulsidi3adddi): Likewise.
16208         (mulsidi3): Likewise.
16209         (*mulsidi3_nov6): Likewise.
16210         (umulsidi3): Likewise.
16211         (umulsidi3_nov6): Likewise.
16212         (umaddsidi4): Likewise.
16213         (*umulsidi3adddi): Likewise.
16214         (smulsi3_highpart): Likewise.
16215         (*smulsi3_highpart_nov6): Likewise.
16216         (umulsi3_highpart): Likewise.
16217         (*umulsi3_highpart_nov6): Likewise.
16218         * config/arm/arm.h (arm_arch3m): Delete.
16219         * config/arm/arm.c (arm_arch3m): Delete.
16220         (arm_option_override_internal): Update armv3-related comment.
16221         (arm_configure_build_target): Delete use of isa_bit_mode32.
16222         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
16223         (arm_rtx_costs_internal): Delete check of arm_arch3m.
16224         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
16225         (mulsa3): Likewise.
16226         (mulusa3): Likewise.
16227         * config/arm/arm-protos.h (arm_arch3m): Delete.
16228         * config/arm/arm-tables.opt: Regenerate.
16229         * config/arm/arm-tune.md: Likewise.
16230         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
16231         deleted architectures.
16233 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16235         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
16236         (armv5t, armv5te): New features.
16237         (ARMv5, ARMv5e): Delete fgroups.
16238         (ARMv5t, ARMv5te): Adjust for above changes.
16239         (ARMv6m): Likewise.
16240         (armv5, armv5e): Delete arches.
16241         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
16242         arm_arch5.
16243         (*call_reg_arm): Likewise.
16244         (*call_value_reg_armv5): Likewise.
16245         (*call_value_reg_arm): Likewise.
16246         (*call_symbol): Likewise.
16247         (*call_value_symbol): Likewise.
16248         (*sibcall_insn): Likewise.
16249         (*sibcall_value_insn): Likewise.
16250         (clzsi2): Likewise.
16251         (prefetch): Likewise.
16252         (define_split and define_peephole2 dependent on arm_arch5):
16253         Likewise.
16254         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
16255         arm_arch5e.
16256         (TARGET_ARM_QBIT): Likewise.
16257         (TARGET_DSP_MULTIPLY): Likewise.
16258         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
16259         (arm_arch5, arm_arch5e): Delete.
16260         (arm_arch5t, arm_arch5te): Declare.
16261         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
16262         (arm_arch5t): Declare.
16263         (arm_option_reconfigure_globals): Update for the above.
16264         (arm_options_perform_arch_sanity_checks): Update comment, replace
16265         use of arm_arch5 with arm_arch5t.
16266         (use_return_insn): Likewise.
16267         (arm_emit_call_insn): Likewise.
16268         (output_return_instruction): Likewise.
16269         (arm_final_prescan_insn): Likewise.
16270         (arm_coproc_builtin_available): Likewise.
16271         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
16272         arm_arch5e with arm_arch5t and arm_arch5te.
16273         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
16274         (arm_arch5t, arm_arch5te): Declare.
16275         * config/arm/arm-tables.opt: Regenerate.
16276         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
16277         * config/arm/t-multilib: Likewise.
16278         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
16279         instead of arm_arch5.
16280         (*call_reg_thumb1): Likewise.
16281         (*call_value_reg_thumb1_v5): Likewise.
16282         (*call_value_reg_thumb1): Likewise.
16283         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
16284         unreachable path.
16285         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
16287 2018-05-18  Martin Liska  <mliska@suse.cz>
16289         PR gcov-profile/84846
16290         * doc/gcov.texi: Document -t option of gcov tool.
16292 2018-05-18  Martin Liska  <mliska@suse.cz>
16294         PR gcov-profile/84846
16295         * gcov.c (print_usage): Add new -t option.
16296         (process_args): Handle the option.
16297         (generate_results): Use stdout as output when requested by
16298         the option.
16300 2018-05-18  Martin Liska  <mliska@suse.cz>
16302         PR gcov-profile/84846
16303         * coverage.c (coverage_init): Write PWD to .gcno file.
16304         * doc/gcov.texi: Document how working directory is printed.
16305         * gcov-dump.c (dump_gcov_file): Print PWD.
16306         * gcov.c (output_intermediate_file): Likewise.
16307         (read_graph_file): Read PWD string.
16308         (output_lines): Print PWD.
16310 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16312         PR middle-end/85817
16313         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
16314         for retval and return false if all args to phi are zero.
16316 2018-05-18  Richard Biener  <rguenther@suse.de>
16318         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
16319         method.
16320         (evrp_dom_walker::before_dom_children): Call it.
16322 2018-05-18  Richard Biener  <rguenther@suse.de>
16324         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
16325         results when processing array refs with variable index.
16327 2018-05-18  Toon Moene  <toon@moene.org>
16329         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
16330         directly after that of -floop-interchange. Indicate that both
16331         options are enabled by default when specifying -O3.
16333 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16335         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
16336         iterator.  Delete separate integer-mode vec_set<mode> expander.
16337         (aarch64_simd_vec_setv2di): Delete.
16338         (vec_setv2di): Delete.
16339         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
16340         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
16341         the "w, r" alternative.
16343 2018-05-18  Martin Liska  <mliska@suse.cz>
16345         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
16346         * tree-pass.h (make_pass_lower_switch_O0): New function.
16347         * tree-switch-conversion.c (node_has_low_bound): Remove.
16348         (node_has_high_bound): Likewise.
16349         (node_is_bounded): Likewise.
16350         (class pass_lower_switch): Make it a template type and create
16351         two instances.
16352         (pass_lower_switch::execute): Add template argument.
16353         (make_pass_lower_switch): New function.
16354         (make_pass_lower_switch_O0): New function.
16355         (do_jump_if_equal): Remove.
16356         (emit_case_nodes): Simplify to just handle all 3 cases and leave
16357         all the hard work to tree optimization passes.
16359 2018-05-18  Martin Liska  <mliska@suse.cz>
16361         * dbgcnt.c (limit_low): Renamed from limit.
16362         (limit_high): New variable.
16363         (dbg_cnt_is_enabled): Check for upper limit.
16364         (dbg_cnt): Adjust dumping.
16365         (dbg_cnt_set_limit_by_index): Add new argument for high
16366         value.
16367         (dbg_cnt_set_limit_by_name): Likewise.
16368         (dbg_cnt_process_single_pair): Parse new format.
16369         (dbg_cnt_process_opt): Use strtok.
16370         (dbg_cnt_list_all_counters): Remove 'value' and add
16371         'limit_high'.
16372         * doc/invoke.texi: Document changes.
16374 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
16376         * doc/sourcebuild.texi (scalar_all_fma): Document.
16377         * tree.def (FMA_EXPR): Delete.
16378         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
16379         * internal-fn.c (ternary_direct): New macro.
16380         (expand_ternary_optab_fn): Likewise.
16381         (direct_ternary_optab_supported_p): Likewise.
16382         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
16383         * builtins.c (fold_builtin_fma): Delete.
16384         (fold_builtin_3): Don't call it.
16385         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
16386         * expr.c (expand_expr_real_2): Likewise.
16387         * fold-const.c (operand_equal_p): Likewise.
16388         (fold_ternary_loc): Likewise.
16389         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
16390         * gimple.c (DEFTREECODE): Likewise.
16391         * gimplify.c (gimplify_expr): Likewise.
16392         * optabs-tree.c (optab_for_tree_code): Likewise.
16393         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
16394         * tree-eh.c (operation_could_trap_p): Likewise.
16395         (stmt_could_throw_1_p): Likewise.
16396         * tree-inline.c (estimate_operator_cost): Likewise.
16397         * tree-pretty-print.c (dump_generic_node): Likewise.
16398         (op_code_prio): Likewise.
16399         * tree-ssa-loop-im.c (stmt_cost): Likewise.
16400         * tree-ssa-operands.c (get_expr_operands): Likewise.
16401         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
16402         * fold-const-call.h (fold_fma): Delete.
16403         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
16404         CFN_FNMA and CFN_FNMS.
16405         (fold_fma): Delete.
16406         * genmatch.c (combined_fn): New enum.
16407         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
16408         (commutative_op): New function.
16409         (commutate): Use it.  Handle more than 2 operands.
16410         (dt_operand::gen_gimple_expr): Use commutative_op.
16411         (parser::parse_expr): Allow :c to be used with non-binary
16412         operators if the commutative operand is known.
16413         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
16414         CFN_FMS, CFN_FNMA and CFN_FNMS.
16415         (backprop::process_assign_use): Remove FMA_EXPR handling.
16416         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
16417         (gen_hsa_fma): New function.
16418         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
16419         IFN_FNMA and IFN_FNMS.
16420         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
16421         * gimple-fold.h (follow_all_ssa_edges): Declare.
16422         * gimple-fold.c (follow_all_ssa_edges): New function.
16423         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
16424         gimple_build interface and use follow_all_ssa_edges to fold the result.
16425         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
16426         instead of checking for optabs directly.
16427         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
16428         rather than FMA_EXPRs.
16429         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
16430         call to IFN_FMA instead of an FMA_EXPR.
16432 2018-05-17  Jim Wilson  <jimw@sifive.com>
16434         * expr.c (do_tablejump): When converting index to Pmode, if we have a
16435         sign extended promoted subreg, and the range does not have the sign bit
16436         set, then do a sign extend.
16438         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
16439         test, check for sign extended subreg and/or constant operands, and
16440         do a sign extend in that case.
16442 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
16444         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
16445         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
16446         Add untyped.
16447         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
16448         Change logics_shift_reg to logics_shift_imm.
16449         (thunderx2t99_fp_loadpair_basic): Delete.
16450         (thunderx2t99_fp_storepair_basic): Delete.
16451         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
16452         (thunderx2t99_asimd_polynomial): Delete.
16453         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
16454         and neon_fp_mul_d_scalar_q.
16455         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
16456         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
16457         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
16458         (thunderx2t99_asimd_lut): Add missing tbl types.
16459         (thunderx2t99_asimd_ext): Delete.
16460         (thunderx2t99_asimd_load1_1_mult): Delete.
16461         (thunderx2t99_asimd_load1_2_mult): Delete.
16462         (thunderx2t99_asimd_load1_ldp): New.
16463         (thunderx2t99_asimd_load1): New.
16464         (thunderx2t99_asimd_load2): Add missing *load2* types.
16465         (thunderx2t99_asimd_load3): New.
16466         (thunderx2t99_asimd_load4): New.
16467         (thunderx2t99_asimd_store1_1_mult): Delete.
16468         (thunderx2t99_asimd_store1_2_mult): Delete.
16469         (thunderx2t99_asimd_store2_mult): Delete.
16470         (thunderx2t99_asimd_store2_onelane): Delete.
16471         (thunderx2t99_asimd_store_stp): New.
16472         (thunderx2t99_asimd_store1): New.
16473         (thunderx2t99_asimd_store2): New.
16474         (thunderx2t99_asimd_store3): New.
16475         (thunderx2t99_asimd_store4): New.
16477 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
16479         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
16480         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
16482 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
16483             Segher Boessenkool  <segher@kernel.crashing.org>
16485         PR target/85698
16486         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
16487         operand.
16489 2018-05-17  Richard Biener  <rguenther@suse.de>
16491         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
16492         for pruning loop and prune defs feeding only already visited PHIs.
16494 2018-05-17  Richard Biener  <rguenther@suse.de>
16496         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
16498 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
16499             Richard Biener  <rguenther@suse.de>
16501         PR tree-optimization/85793
16502         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
16503         for VMAT_ELEMENTWISE.
16505 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
16507         * internal-fn.h (lookup_internal_fn): Declare
16508         * internal-fn.c (lookup_internal_fn): New function.
16509         * gimple.c (gimple_build_call_from_tree): Handle calls to
16510         internal functions.
16511         * gimple-pretty-print.c (dump_gimple_call): Print "." before
16512         internal function names.
16513         * tree-pretty-print.c (dump_generic_node): Likewise.
16514         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
16516 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
16518         * gimple-fold.h (gimple_build): Make the function forms take
16519         combined_fn rather than built_in_function.
16520         (gimple_simplify): Likewise.
16521         * gimple-match-head.c (gimple_simplify): Likewise.
16522         * gimple-fold.c (gimple_build): Likewise.
16523         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
16524         rather than gimple_build_call_internal.
16525         (get_initial_defs_for_reduction): Likewise.
16526         (vect_create_epilog_for_reduction): Likewise.
16527         (vectorizable_live_operation): Likewise.
16529 2018-05-17  Martin Liska  <mliska@suse.cz>
16531         * gimple-ssa-sprintf.c (format_directive): Do not use
16532         space in between 'G_' and '('.
16534 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
16536         PR target/85323
16537         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
16538         even if the mask is not all ones.
16540         PR target/85323
16541         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
16542         vector.
16543         (ix86_gimple_fold_builtin): Likewise.
16545         PR target/85323
16546         * config/i386/i386.c: Include tree-vector-builder.h.
16547         (ix86_vector_shift_count): New function.
16548         (ix86_fold_builtin): Fold shift builtins by scalar count.
16549         (ix86_gimple_fold_builtin): Likewise.
16551         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
16552         _mm512_setzero): New intrinsics.
16554 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
16555             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16557         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
16558         code generation for cases where splatting a value is not useful.
16559         * simplify-rtx.c (simplify_ternary_operation): Simplify
16560         vec_merge across a vec_duplicate and a paradoxical subreg forming
16561         a vector mode to a vec_concat.
16563 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
16565         * config.gcc: Support "goldmont-plus".
16566         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
16567         "goldmont-plus".
16568         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16569         PROCESSOR_GOLDMONT_PLUS.
16570         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
16571         (processor_target_table): Add "goldmont-plus".
16572         (PTA_GOLDMONT_PLUS): Define.
16573         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
16574         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
16575         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
16576         (fold_builtin_cpu): Add "goldmont-plus".
16577         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
16578         (ix86_option_override_internal): Add "goldmont-plus".
16579         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
16580         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
16581         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
16582         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
16584 2018-05-17  Richard Biener  <rguenther@suse.de>
16586         PR tree-optimization/85757
16587         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
16588         remove defs that only feed that PHI from further processing.
16590 2018-05-16  Jim Wilson  <jimw@sifive.com>
16592         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
16593         asterisk to name.
16594         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
16596 2018-05-16  Mark Wielaard  <mark@klomp.org>
16598         * dwarf2out.c (count_index_strings): New function.
16599         (output_indirect_strings): Call count_index_strings and generate
16600         header for dwarf_version >= 5.
16602 2018-05-16  Mark Wielaard  <mark@klomp.org>
16604         * dwarf2out.c (dwarf_FORM): New function.
16605         (set_indirect_string): Use dwarf_FORM.
16606         (reset_indirect_string): Likewise.
16607         (size_of_die): Likewise.
16608         (value_format): Likewise.
16609         (output_die): Likewise.
16610         (add_skeleton_AT_string): Likewise.
16611         (output_macinfo_op): Likewise.
16612         (index_string): Likewise.
16613         (output_index_string_offset): Likewise.
16614         (output_index_string): Likewise.
16615         (count_index_strings): Likewise.
16617 2018-05-16  Carl Love  <cel@us.ibm.com>
16619         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
16620         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
16622 2018-05-16  Martin Jambor  <mjambor@suse.cz>
16624         * ipa-prop.c (ipa_free_all_edge_args): Remove.
16625         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
16627 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
16629         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
16630         (fnma<mode>4): Likewise.
16631         (fms<mode>4): Likewise.
16632         (fnms<mode>4): Likewise.
16633         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
16634         (aarch64_fnma<mode>4): Likewise.
16635         (aarch64_fms<mode>4): Likewise.
16636         (aarch64_fnms<mode>4): Likewise.
16637         (aarch64_fnmadd<mode>4): Likewise.
16639 2018-05-16  Jason Merrill  <jason@redhat.com>
16641         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
16643 2018-05-16  Richard Biener  <rguenther@suse.de>
16645         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
16646         (dump_stmt_cost): Declare.
16647         (add_stmt_cost): Dump cost we add.
16648         (add_stmt_costs): New function.
16649         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
16650         No longer exported.
16651         (vect_analyze_stmt): Adjust prototype.
16652         (vectorizable_condition): Likewise.
16653         (vectorizable_live_operation): Likewise.
16654         (vectorizable_reduction): Likewise.
16655         (vectorizable_induction): Likewise.
16656         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
16657         cost vector to pass to vectorizable_ and record afterwards.
16658         (vect_model_reduction_cost): Take cost vector argument and adjust.
16659         (vect_model_induction_cost): Likewise.
16660         (vectorizable_reduction): Likewise.
16661         (vectorizable_induction): Likewise.
16662         (vectorizable_live_operation): Likewise.
16663         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
16664         SLP_TREE_NUMBER_OF_VEC_STMTS.
16665         (vect_analyze_slp_cost_1): Remove.
16666         (vect_analyze_slp_cost): Likewise.
16667         (vect_slp_analyze_node_operations): Take visited args and
16668         a target cost vector.  Avoid processing already visited stmt sets.
16669         (vect_slp_analyze_operations): Use a local cost vector to gather
16670         costs and register those of non-discarded instances.
16671         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
16672         (vect_schedule_slp_instance): Remove copying of
16673         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
16674         zero.
16675         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
16676         adding cost.  Record cost entry location.
16677         (vect_prologue_cost_for_slp_op): Function to compute cost of
16678         a constant or invariant generated for SLP vect in the prologue,
16679         split out from vect_analyze_slp_cost_1.
16680         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
16681         (vect_model_promotion_demotion_cost): Likewise.
16682         (vect_model_store_cost): Likewise, make static.
16683         (vect_model_load_cost): Likewise.
16684         (vectorizable_bswap): Add cost vector arg and adjust.
16685         (vectorizable_call): Likewise.
16686         (vectorizable_simd_clone_call): Likewise.
16687         (vectorizable_conversion): Likewise.
16688         (vectorizable_assignment): Likewise.
16689         (vectorizable_shift): Likewise.
16690         (vectorizable_operation): Likewise.
16691         (vectorizable_store): Likewise.
16692         (vectorizable_load): Likewise.
16693         (vectorizable_condition): Likewise.
16694         (vectorizable_comparison): Likewise.
16695         (can_vectorize_live_stmts): Likewise.
16696         (vect_analyze_stmt): Likewise.
16697         (vect_transform_stmt): Adjust calls to vectorizable_*.
16698         * tree-vectorizer.c: Include gimple-pretty-print.h.
16699         (dump_stmt_cost): New function.
16701 2018-05-16  Richard Biener  <rguenther@suse.de>
16703         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
16704         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
16705         * tree-ssa-dse.c: Include tree-ssa-loop.h.
16706         (check_name): New callback.
16707         (dse_classify_store): Track cycles via a visited bitmap of PHI
16708         defs and simplify handling of in-loop and across loop dead stores
16709         and properly fail for loop-variant refs.  Handle byte-tracking with
16710         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
16711         limiting the walk.
16713 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
16715         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
16716         (vect_get_mask_type_for_stmt): Likewise.
16717         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
16718         split out from...
16719         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
16720         to determine the statement's vector type and the vector type that
16721         should be used for calculating nunits.  Deal with cases in which
16722         the type has to be deferred.
16723         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
16724         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
16725         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
16726         (vect_determine_vf_for_stmt): New functions, split out from...
16727         (vect_determine_vectorization_factor): ...here.
16728         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
16729         (vect_get_mask_type_for_stmt): New functions, split out from
16730         vect_determine_vectorization_factor.
16732 2018-05-16  Richard Biener  <rguenther@suse.de>
16734         * tree-cfg.c (verify_gimple_assign_ternary): Properly
16735         verify the [VEC_]COND_EXPR embedded comparison.
16737 2018-05-15  Martin Sebor  <msebor@redhat.com>
16739         PR tree-optimization/85753
16740         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
16741         RECORD_TYPE in addition to ARRAY_TYPE.
16743 2018-05-15  Martin Sebor  <msebor@redhat.com>
16745         PR middle-end/85643
16746         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
16748 2018-05-15  Richard Biener  <rguenther@suse.de>
16750         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
16751         add by_clobber_p one.  Change algorithm to collect all defs
16752         representing uses we need to walk and try reducing them to
16753         a single one before failing.
16754         (dse_dom_walker::dse_optimize_stmt): Adjust.
16756 2018-05-13  Mark Wielaard  <mark@klomp.org>
16758         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
16759         (size_of_loc_descr): Likewise.
16760         (output_loc_operands): Likewise.
16761         (output_loc_operands_raw): Likewise.
16762         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
16763         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
16764         (hash_loc_operands): Likewise.
16765         (compare_loc_operands): Likewise.
16767 2018-05-14  Mark Wielaard  <mark@klomp.org>
16769         * dwarf2out.c (count_index_addrs): New function.
16770         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
16772 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16774         PR tree-optimization/83648
16775         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
16776         return value as malloc candidate.
16778 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16780         PR ipa/85734
16781         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
16782         param as true in call to suggest_attribute.
16784 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
16786         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
16787         -mreadonly-in-sdata.
16789 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16791         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
16792         New pattern.
16793         (aarch64_crypto_aesd_fused): Likewise.
16795 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
16797         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
16798         (movsi_aarch64): Likewise.
16799         (load_pairsi): Likewise.
16800         (load_pairdi): Likewise.
16801         (store_pairsi): Likewise.
16802         (store_pairdi): Likewise.
16803         (load_pairsf): Likewise.
16804         (load_pairdf): Likewise.
16805         (store_pairsf): Likewise.
16806         (store_pairdf): Likewise.
16807         (zero_extend): Likewise.
16808         (trunc): Swap alternatives.
16809         (fcvt_target): Add '?' to prefer w over r.
16811 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
16813         PR target/85756
16814         * config/i386/i386.md: Disallow non-commutative arithmetics in
16815         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
16816         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
16817         in the peephole2 before it.
16819 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
16821         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
16822         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
16823         (ix86_handle_option): Handle -mcldemote.
16824         * config.gcc: New header.
16825         * config/i386/cldemoteintrin.h: New file.
16826         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
16827         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
16828         -mcldemote.
16829         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16830         OPTION_MASK_ISA_CLDEMOTE.
16831         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
16832         (ix86_valid_target_attribute_inner_p): Ditto.
16833         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
16834         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
16835         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
16836         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
16837         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
16838         (cldemote): New.
16839         * config/i386/i386.opt: Add -mcldemote.
16840         * config/i386/x86intrin.h: New header.
16841         * doc/invoke.texi: Add -mcldemote.
16843 2018-05-14  Richard Biener  <rguenther@suse.de>
16845         * doc/match-and-simplify.texi: Adjust :s documentation.
16847 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
16849         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
16850         intended memcpy size.
16851         (REORDER_45): Likewise.
16853 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
16855         * sort.cc: New file.
16856         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
16857         * vec.c (qsort_chk): Use gcc_qsort.
16858         * Makefile.in (OBJS-libcommon): Add sort.o.
16859         (build/sort.o): New target.  Use it...
16860         (BUILD_RTL): ... here, and...
16861         (build/gencfn-macros): ... here, and...
16862         (build/genmatch): ... here.
16864 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
16865             Chung-Ju Wu  <jasonwucj@gmail.com>
16867         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
16868         * config/nds32/nds32-graywolf.md: New file.
16869         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
16870         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
16871         pipeline.
16872         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
16873         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
16874         * config/nds32/nds32.md (pipeline_model): Add graywolf.
16875         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
16876         * config/nds32/pipelines.md: Include n15 settings.
16878 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
16879             Chung-Ju Wu  <jasonwucj@gmail.com>
16881         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
16882         * config/nds32/nds32-n13.md: New file.
16883         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
16884         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
16885         pipeline.
16886         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
16887         * config/nds32/nds32.md (pipeline_model): Add n13.
16888         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
16889         * config/nds32/pipelines.md: Include n13 settings.
16891 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
16892             Chung-Ju Wu  <jasonwucj@gmail.com>
16894         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
16895         * config/nds32/nds32-n10.md: New file.
16896         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
16897         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
16898         pipeline.
16899         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
16900         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
16901         * config/nds32/nds32.md (pipeline_model): Add n10.
16902         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
16903         * config/nds32/pipelines.md: Include n10 settings.
16905 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
16906             Kito Cheng  <kito.cheng@gmail.com>
16907             Chung-Ju Wu  <jasonwucj@gmail.com>
16909         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
16910         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16911         Add enum values for DSP extension instructions.
16912         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
16913         New constraints.
16914         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
16915         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
16916         New code iterators.
16917         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
16918         * config/nds32/nds32-dspext.md: New file for DSP implementation.
16919         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
16920         * config/nds32/nds32-intrinsic.md: Likewise.
16921         * config/nds32/nds32_intrinsic.h: Likewise.
16922         * config/nds32/nds32-md-auxiliary.c: Likewise.
16923         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
16924         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
16925         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
16926         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
16927         * config/nds32/nds32-protos.h: New declarations for DSP extension.
16928         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
16929         TYPE_DMAC in switch statement.
16930         * config/nds32/nds32.c: New checking and implementation for DSP
16931         extension instructions.
16932         * config/nds32/nds32.h: Likewise.
16933         * config/nds32/nds32.md: Likewise.
16934         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
16935         * config/nds32/predicates.md: Implement new predicates for DSP
16936         extension.
16938 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
16940         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
16941         Reformat alternatives and attributes so it is easier to identify
16942         which constraints/attributes go with which instruction.
16943         (mov<mode>_hardfloat32, FMOVE64): Likewise.
16944         (mov<mode>_softfloat32, FMOVE64): Likewise.
16945         (mov<mode>_hardfloat64, FMOVE64): Likewise.
16946         (mov<mode>_softfloat64, FMOVE64): Likewise.
16948 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16950         * doc/extend.texi (PowerPC Built-in Functions): Rename this
16951         subsection.
16952         (Basic PowerPC Built-in Functions): The new name of the
16953         subsection previously known as "PowerPC Built-in Functions".
16954         (Basic PowerPC Built-in Functions Available on all Configurations):
16955         New subsubsection.
16956         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
16957         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
16958         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
16959         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
16961 2018-05-11  Martin Jambor  <mjambor@suse.cz>
16963         PR ipa/85655
16964         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
16965         single const.
16967 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
16969         PR target/85733
16970         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
16972 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
16974         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
16975         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
16976         (ix86_handle_option): Handle -mwaitpkg.
16977         * config.gcc: New header.
16978         * config/i386/cpuid.h (bit_WAITPKG): New bit.
16979         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
16980         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
16981         function type.
16982         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16983         OPTION_MASK_ISA_WAITPKG.
16984         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
16985         (ix86_option_override_internal): Add PTA_WAITPKG.
16986         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
16987         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
16988         IX86_BUILTIN_TPAUSE.
16989         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
16990         __builtin_ia32_umwait and __builtin_ia32_tpause.
16991         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
16992         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
16993         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
16994         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
16995         UNSPECV_TPAUSE): New.
16996         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
16997         * config/i386/i386.opt: Add -mwaitpkg.
16998         * config/i386/waitpkgintrin.h: New file.
16999         * config/i386/x86intrin.h: New header.
17000         * doc/invoke.texi: Add -mwaitpkg.
17002 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
17004         PR target/85606
17005         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
17006         equivalent.
17007         (cortex-m0): Use armv6s-m isa.
17008         (cortex-m0plus): Likewise.
17009         (cortex-m1): Likewise.
17010         (cortex-m0.small-multiply): Likewise.
17011         (cortex-m0plus.small-multiply): Likewise.
17012         (cortex-m1.small-multiply): Likewise.
17014 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
17015             Jakub Jelinek  <jakub@redhat.com>
17017         PR tree-optimization/85692
17018         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
17019         source permute as well.
17021 2018-05-11  Martin Liska  <mliska@suse.cz>
17023         PR sanitizer/85556
17024         * doc/extend.texi: Document LLVM style format for no_sanitize
17025         attribute.
17027 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
17029         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
17030         mode_supports_vsx_dform_quad to mode_supports_dq_form.
17031         (mode_supports_vsx_dform_quad): Likewise.
17032         (mode_supports_vmx_dform): Move these functions to be next to the
17033         other mode_supports functions.
17034         (mode_supports_dq_form): Likewise.
17035         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
17036         mode_supports_dq_form.
17037         (reg_offset_addressing_ok_p): Likewise.
17038         (offsettable_ok_by_alignment): Likewise.
17039         (rs6000_legitimate_offset_address_p): Likewise.
17040         (legitimate_lo_sum_address_p): Likewise.
17041         (rs6000_legitimize_address): Likewise.
17042         (rs6000_legitimize_reload_address): Likewise.
17043         (rs6000_secondary_reload_inner): Likewise.
17044         (rs6000_preferred_reload_class): Likewise.
17045         (rs6000_output_move_128bit): Likewise.
17047 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
17049         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
17050         Generate SImode target register for null target.
17051         <case IX86_BUILTIN_XGETBV>: Ditto.
17052         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
17053         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
17055 2018-05-10  Carl Love  <cel@us.ibm.com>
17057         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
17058         dcbtt and dcbtstt if operands[2] is 0.
17060 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
17062         PR target/85693
17063         * config/i386/sse.md (usadv64qi): New expander.
17065 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
17067         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
17068         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
17069         -maltivec=be support.
17070         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
17071         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
17072         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
17073         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
17074         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
17075         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
17076         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
17077         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
17078         altivec_vsumsws): Adjust.
17079         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
17080         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
17081         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
17082         support.
17083         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
17084         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
17085         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
17086         (altivec_lve<VI_char>x): Delete expand.
17087         (*altivec_lve<VI_char>x_internal): Rename to...
17088         (altivec_lve<VI_char>x): ... this.
17089         (altivec_lvxl_<mode>): Delete expand.
17090         (*altivec_lvxl_<mode>_internal): Rename to ...
17091         (altivec_lvxl_<mode>): ... this.
17092         (altivec_stvxl_<mode>): Delete expand.
17093         (*altivec_stvxl_<mode>_internal): Rename to ...
17094         (altivec_stvxl_<mode>): ... this.
17095         (altivec_stve<VI_char>x): Delete expand.
17096         (*altivec_stve<VI_char>x_internal): Rename to ...
17097         (altivec_stve<VI_char>x): ... this.
17098         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
17099         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
17100         reduc_plus_scal_<mode>): Adjust.
17101         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
17102         comment.
17103         (rs6000_cpu_cpp_builtins): Adjust.
17104         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
17105         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
17106         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
17107         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
17108         -maltivec=be support.
17109         (rs6000_split_vec_extract_var): Adjust.
17110         (rs6000_split_v4si_init): Adjust.
17111         (swap_selector_for_mode): Delete.
17112         (altivec_expand_lvx_be, altivec_expand_stvx_be,
17113         altivec_expand_stvex_be): Delete.
17114         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
17115         -maltivec=be support.
17116         (rs6000_gimple_fold_builtin): Ditto.
17117         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
17118         Adjust.
17119         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
17120         (TARGET_DIRECT_MOVE_64BIT): Adjust.
17121         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
17122         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
17123         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
17124         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
17125         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
17126         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
17127         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
17128         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
17129         anonymous split): Adjust.
17130         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
17131         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
17133 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
17135         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
17136         when --with-gxx-include-dir is also specified.
17137         * configure: Regenerate.
17139 2018-05-09  Jim Wilson  <jimw@sifive.com>
17141         PR target/84797
17142         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
17143         * config/riscv/t-withmultilib: New.
17144         * config/riscv/withmultilib.h: New.
17145         * doc/install.texi: Document RISC-V --with-multilib-list support.
17147 2018-05-09  Richard Biener  <rguenther@suse.de>
17149         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
17150         vector.
17151         (vect_bb_vectorization_profitable_p): Adjust.  Compute
17152         actual scalar cost using the cost vector and the add_stmt_cost
17153         machinery.
17155 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17157         PR rtl-optimization/85645
17158         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
17159         in the REG_CFA_REGISTER note for LR, don't leave it empty.
17161 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17163         PR rtl-optimization/85645
17164         * shrink-wrap.c (spread_components): Return a boolean saying if
17165         anything was changed.
17166         (try_shrink_wrapping_separate): Iterate spread_components until
17167         nothing changes anymore.
17169 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17171         PR rtl-optimization/85645
17172         * regrename.c (build_def_use): Also kill the chains that include the
17173         destination of a REG_CFA_REGISTER note.
17175 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
17177         PR rtl-optimization/85645
17178         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
17179         insn that has a REG_CFA_REGISTER note.
17181 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
17183         * cfgexpand.c (expand_clobber): New function.
17184         (expand_gimple_stmt_1): Use it.
17185         * tree-vect-stmts.c (vect_clobber_variable): New function,
17186         split out from...
17187         (vectorizable_simd_clone_call): ...here.
17188         (vectorizable_store): Emit a clobber either side of an
17189         IFN_STORE_LANES sequence.
17190         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
17192 2018-05-09  Tom de Vries  <tom@codesourcery.com>
17194         PR target/85626
17195         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
17196         (define_insn "trap_if_false"): Add exit after trap.
17198 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
17200         PR rtl-optimization/85638
17201         * bb-reorder.c: Include common/common-target.h.
17202         (create_forwarder_block): New function extracted from...
17203         (fix_up_crossing_landing_pad): ...here.  Rename into...
17204         (dw2_fix_up_crossing_landing_pad): ...this.
17205         (sjlj_fix_up_crossing_landing_pad): New function.
17206         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
17207         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
17208         from both partitions and exit the loop after one iteration.
17210 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17212         Revert:
17213         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17214         subsection.
17215         (Basic PowerPC Built-in Functions): The new name of the
17216         subsection previously known as "PowerPC Built-in Functions".
17217         (Basic PowerPC Built-in Functions Available on all Configurations):
17218         New subsubsection.
17219         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
17220         subsubsection.
17221         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
17222         subsubsection.
17223         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
17224         subsubsection.
17225         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
17226         subsubsection.
17228 2018-05-08  Jim Wilson  <jimw@sifive.com>
17230         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
17231         (LD_EMUL_SUFFIX): New.
17232         (LINK_SPEC): Use it.
17234 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17236         * doc/extend.texi (PowerPC Built-in Functions): Rename this
17237         subsection.
17238         (Basic PowerPC Built-in Functions): The new name of the
17239         subsection previously known as "PowerPC Built-in Functions".
17240         (Basic PowerPC Built-in Functions Available on all Configurations):
17241         New subsubsection.
17242         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
17243         subsubsection.
17244         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
17245         subsubsection.
17246         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
17247         subsubsection.
17248         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
17249         subsubsection.
17251 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
17253         PR target/85683
17254         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
17255         after cmpelim optimization.
17257 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
17259         * config.gcc: Support "goldmont".
17260         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
17261         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17262         PROCESSOR_GOLDMONT.
17263         * config/i386/i386.c (m_GOLDMONT): Define.
17264         (processor_target_table): Add "goldmont".
17265         (PTA_GOLDMONT): Define.
17266         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
17267         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
17268         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
17269         (fold_builtin_cpu): Add "goldmont".
17270         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
17271         (ix86_option_override_internal): Add "goldmont".
17272         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
17273         (processor_type): Add PROCESSOR_GOLDMONT.
17274         * config/i386/i386.md: Add CPU "glm".
17275         * config/i386/glm.md: New file.
17276         * config/i386/x86-tune.def: Add m_GOLDMONT.
17277         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
17279 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
17281         PR target/85572
17282         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
17283         E_V4DImode.
17284         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
17285         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
17286         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
17288         PR target/85317
17289         * config/i386/i386.c (ix86_fold_builtin): Handle
17290         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
17292         PR target/85480
17293         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
17294         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
17296 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
17298         PR target/85658
17299         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
17300         (check_arch): Likewise.
17301         (check_fpu): Return the result rather than printing it.
17302         (end arch): Fix operator precedence.
17303         (end cpu): Likewise.
17304         (END): Print the result from check_fpu.
17306 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
17307             Alan Hayward  <alan.hayward@arm.com>
17308             David Sherwood  <david.sherwood@arm.com>
17310         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
17311         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
17312         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
17313         (*fcmuo<mode>_and): New patterns.
17315 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
17317         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
17318         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
17319         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
17320         (cmp_op, sve_imm_con): New code attributes.
17321         (SVE_COND_INT_CMP, imm_con): Delete.
17322         (cmp_op): Remove above unspecs from int attribute.
17323         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
17324         to...
17325         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
17326         comparison-specific unspecs.
17327         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
17328         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
17329         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
17330         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
17331         (*vec_fcm<cmp_op><mode>): Rename to...
17332         (*fcm<cmp_op><mode>): ...this and adjust likewise.
17333         (*vec_fcmuo<mode>): Rename to...
17334         (*fcmuo<mode>): ...this and adjust likewise.
17335         (*pred_fcm<cmp_op><mode>): New pattern.
17336         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
17337         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
17338         functions.
17339         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
17340         and UNORDERED.
17341         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
17342         (aarch64_emit_sve_predicated_cond): New function.
17343         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
17344         (aarch64_emit_unspec_cond_or): Replace with...
17345         (aarch64_emit_sve_or_conds): ...this new function.  Use
17346         aarch64_emit_sve_ptrue_op for the individual comparisons and
17347         aarch64_emit_binop to OR them together.
17348         (aarch64_emit_inverted_unspec_cond): Replace with...
17349         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
17350         aarch64_emit_sve_ptrue_op for the comparison and
17351         aarch64_emit_unop to invert the result.
17352         (aarch64_expand_sve_vec_cmp_float): Update after the above
17353         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
17355 2018-05-07  Nathan Sidwell  <nathan@acm.org>
17357         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
17358         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
17359         (Backwards Compatibility): Likewise.
17361 2018-05-07  Luis Machado  <luis.machado@linaro.org>
17363         PR bootstrap/85681
17364         Revert:
17365         2018-05-07  Luis Machado  <luis.machado@linaro.org>
17367         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17368         <prefetch_dynamic_strides>: New const bool field.
17369         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17370         prefetch_dynamic_strides.
17371         (exynosm1_prefetch_tune): Likewise.
17372         (thunderxt88_prefetch_tune): Likewise.
17373         (thunderx_prefetch_tune): Likewise.
17374         (thunderx2t99_prefetch_tune): Likewise.
17375         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
17376         to false.
17377         (aarch64_override_options_internal): Update to set
17378         PARAM_PREFETCH_DYNAMIC_STRIDES.
17379         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
17380         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
17381         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
17382         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
17383         prefetch-dynamic-strides setting.
17385         2018-05-07  Luis Machado  <luis.machado@linaro.org>
17387         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17388         <minimum_stride>: New const int field.
17389         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17390         minimum_stride field.
17391         (exynosm1_prefetch_tune): Likewise.
17392         (thunderxt88_prefetch_tune): Likewise.
17393         (thunderx_prefetch_tune): Likewise.
17394         (thunderx2t99_prefetch_tune): Likewise.
17395         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
17396         (aarch64_override_options_internal): Update to set
17397         PARAM_PREFETCH_MINIMUM_STRIDE.
17398         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
17399         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
17400         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
17401         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
17402         stride is constant and is below the minimum stride threshold.
17404 2018-05-07  Luis Machado  <luis.machado@linaro.org>
17406         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
17407         to 512.
17409 2018-05-07  Luis Machado  <luis.machado@linaro.org>
17411         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17412         <prefetch_dynamic_strides>: New const bool field.
17413         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17414         prefetch_dynamic_strides.
17415         (exynosm1_prefetch_tune): Likewise.
17416         (thunderxt88_prefetch_tune): Likewise.
17417         (thunderx_prefetch_tune): Likewise.
17418         (thunderx2t99_prefetch_tune): Likewise.
17419         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
17420         to false.
17421         (aarch64_override_options_internal): Update to set
17422         PARAM_PREFETCH_DYNAMIC_STRIDES.
17423         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
17424         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
17425         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
17426         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
17427         prefetch-dynamic-strides setting.
17429 2018-05-07  Luis Machado  <luis.machado@linaro.org>
17431         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
17432         <minimum_stride>: New const int field.
17433         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
17434         minimum_stride field.
17435         (exynosm1_prefetch_tune): Likewise.
17436         (thunderxt88_prefetch_tune): Likewise.
17437         (thunderx_prefetch_tune): Likewise.
17438         (thunderx2t99_prefetch_tune): Likewise.
17439         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
17440         (aarch64_override_options_internal): Update to set
17441         PARAM_PREFETCH_MINIMUM_STRIDE.
17442         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
17443         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
17444         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
17445         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
17446         stride is constant and is below the minimum stride threshold.
17448 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
17450         PR c++/85659
17451         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
17452         the type is addressable.  Don't force op into register if it has
17453         BLKmode.
17455 2018-05-05  Roland McGrath  <mcgrathr@google.com>
17457         PR other/77609
17458         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
17459         any section for which we don't know a specific type it should have,
17460         regardless of name.  Previously this was done only for the exact
17461         names ".init_array", ".fini_array", and ".preinit_array".
17462         (default_elf_asm_named_section): Add comment about
17463         relationship with default_section_type_flags and SECTION_NOTYPE.
17464         (get_section): Don't consider it a type conflict if one side has
17465         SECTION_NOTYPE and the other doesn't, as long as neither has the
17466         SECTION_BSS et al used in the default_section_type_flags logic.
17468 2018-05-05  Tom de Vries  <tom@codesourcery.com>
17470         PR target/85653
17471         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
17472         (workaround_barsyncs): New function.
17473         (nvptx_reorg): Use workaround_barsyncs.
17474         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
17475         (define_expand "nvptx_membar_cta"): New define_expand.
17476         (define_insn "*nvptx_membar_cta"): New insn.
17478 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
17480         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
17481         To improve optimization opportunities.
17482         * builtin-types.def: The new needed builtin types for the above.
17484 2018-05-04  Richard Biener  <rguenther@suse.de>
17486         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
17487         * gimple-ssa-store-merging.c
17488         (imm_store_chain_info::output_merged_store): Remove redundant create,
17489         release split_store vector contents on failure.
17490         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
17491         scalar stmt vector on cache hit.
17493 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
17495         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
17496         Xilinx FP support.
17497         * config.gcc (powerpc-xilinx-eabi*): Remove.
17498         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
17499         support.
17500         (fusion_addis_mem_combo_load): Ditto.
17501         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
17502         FP support.
17503         (rs6000_cpu_cpp_builtins): Ditto.
17504         * config/rs6000/rs6000-linux.c
17505         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
17506         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
17507         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
17508         support.
17509         (rs6000_setup_reg_addr_masks): Ditto.
17510         (rs6000_init_hard_regno_mode_ok): Ditto.
17511         (rs6000_option_override_internal): Ditto.
17512         (legitimate_lo_sum_address_p): Ditto.
17513         (rs6000_legitimize_address): Ditto.
17514         (rs6000_legitimize_reload_address): Ditto.
17515         (rs6000_legitimate_address_p): Ditto.
17516         (abi_v4_pass_in_fpr): Ditto.
17517         (setup_incoming_varargs): Ditto.
17518         (rs6000_gimplify_va_arg): Ditto.
17519         (rs6000_split_multireg_move): Ditto.
17520         (rs6000_savres_strategy): Ditto.
17521         (rs6000_emit_prologue_components): Ditto.
17522         (rs6000_emit_epilogue_components): Ditto.
17523         (rs6000_emit_prologue): Ditto.
17524         (rs6000_emit_epilogue): Ditto.
17525         (rs6000_elf_file_end): Ditto.
17526         (rs6000_function_value): Ditto.
17527         (rs6000_libcall_value): Ditto.
17528         * config/rs6000/rs6000.h: Ditto.
17529         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
17530         (TARGET_MINMAX): ... this.  New.
17531         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
17532         * config/rs6000/rs6000.md: Remove Xilinx FP support.
17533         (*movsi_internal1_single): Delete.
17534         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
17535         mfpu=, mxilinx-fpu): Delete.
17536         * config/rs6000/singlefp.h: Delete.
17537         * config/rs6000/sysv4.h: Remove Xilinx FP support.
17538         * config/rs6000/t-rs6000: Ditto.
17539         * config/rs6000/t-xilinx: Delete.
17540         * config/rs6000/titan.md: Adjust for fp_type removal.
17541         * config/rs6000/vsx.md: Remove Xilinx FP support.
17542         (VStype_simple): Delete.
17543         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
17544         * config/rs6000/xfpu.h: Delete.
17545         * config/rs6000/xfpu.md: Delete.
17546         * config/rs6000/xilinx.h: Delete.
17547         * config/rs6000/xilinx.opt: Delete.
17548         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
17549         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
17551 2018-05-04  Tom de Vries  <tom@codesourcery.com>
17553         PR libgomp/85639
17554         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
17555         if ignore == 0.
17557 2018-05-04  Richard Biener  <rguenther@suse.de>
17559         PR middle-end/85627
17560         * tree-complex.c (update_complex_assignment): We are always in SSA form.
17561         (expand_complex_div_wide): Likewise.
17562         (expand_complex_operations_1): Likewise.
17563         (expand_complex_libcall): Preserve EH info of the original stmt.
17564         (tree_lower_complex): Handle removed blocks.
17565         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
17566         on complex multiplication and division libcall builtins.
17568 2018-05-04  Richard Biener  <rguenther@suse.de>
17570         PR middle-end/85574
17571         * fold-const.c (negate_expr_p): Restrict negation of operand
17572         zero of a division to when we know that can happen without
17573         overflow.
17574         (fold_negate_expr_1): Likewise.
17576 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
17578         PR libstdc++/85466
17579         * real.h (real_nextafter): Declare.
17580         * real.c (real_nextafter): New function.
17581         * fold-const-call.c (fold_const_nextafter): New function.
17582         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
17583         CASE_CFN_NEXTTOWARD.
17584         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
17585         even when arg1_mode is different from arg0_mode.
17587 2018-05-03  Nathan Sidwell  <nathan@acm.org>
17589         * doc/extend.texi (Deprecated Features): Remove
17590         -ffriend-injection.
17591         (Backwards Compatibility): Likewise.
17592         * doc/invoke.texi (C++ Language Options): Likewise.
17593         (C++ Dialect Options): Likewise.
17595 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
17597         PR target/85530
17598         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
17599         _mm512_mask_mullox_epi64): New intrinsics.
17601 2018-05-03  Tom de Vries  <tom@codesourcery.com>
17603         PR testsuite/85106
17604         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17605         dump files): Add offload-tree.
17607 2018-05-03  Richard Biener  <rguenther@suse.de>
17609         PR tree-optimization/85615
17610         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
17611         to loops not nested in BBs loop father to avoid creating multi-entry
17612         loops.
17614 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17616         PR tree-optimization/70291
17617         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
17618         arguments.  Change return type to tree.  Emit libcall as a new
17619         statement rather than replacing existing one when inplace_p is true.
17620         (expand_complex_multiplication_components): New function.
17621         (expand_complex_multiplication): Expand floating-point complex
17622         multiplication using the above.
17623         (expand_complex_division): Rename inner_type parameter to type.
17624         Update expand_complex_libcall call-site.
17625         (expand_complex_operations_1): Update expand_complex_multiplication
17626         and expand_complex_division call-sites.
17628 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
17630         PR target/85582
17631         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
17632         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
17633         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
17634         the highest significant bit of the shift count mask is clear.  In
17635         check whether and[sq]i3 is needed verify that all significant bits
17636         of the shift count other than the highest are set.
17638 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17640         PR libgomp/82428
17641         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
17642         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
17643         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
17644         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
17645         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
17646         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
17647         __builtin_goacc_parlevel_size.
17649 2018-05-02  Richard Biener  <rguenther@suse.de>
17651         PR tree-optimization/85597
17652         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
17653         do not use split vect_get_vec_defs call but call vect_get_slp_defs
17654         directly.
17656 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17658         PR testsuite/85106
17659         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17660         dump files): Add ltrans-tree.
17662 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17664         PR testsuite/85106
17665         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
17666         dump files): Add wpa-ipa.
17668 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
17670         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
17671         powerpc*-*-linux*paired* target.
17672         * config/rs6000/750cl.h: Delete.
17673         * config/rs6000/paired.h: Delete.
17674         * config/rs6000/paired.md: Delete.
17675         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
17676         float support.
17677         * config/rs6000/rs6000-builtin.def: Remove paired float support.
17678         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
17679         comment.  Remove paired float support.
17680         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
17681         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
17682         VECTOR_PAIRED.
17683         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
17684         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
17685         declarations.
17686         * config/rs6000/rs6000.c: Remove paired float support.
17687         (paired_expand_vector_init, paired_expand_vector_move,
17688         paired_emit_vector_compare, paired_emit_vector_cond_expr,
17689         (paired_expand_lv_builtin, paired_expand_stv_builtin,
17690         paired_expand_builtin, paired_expand_predicate_builtin,
17691         paired_init_builtins): Delete.
17692         * config/rs6000/rs6000.h: Remove paired float support.
17693         * config/rs6000/rs6000.md: Remove paired float support.
17694         (move_from_CR_ov_bit): Delete.
17695         * config/rs6000/rs6000.opt (mpaired): Delete.
17696         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
17697         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
17699 2018-05-02  Richard Biener  <rguenther@suse.de>
17701         PR middle-end/85567
17702         * gimplify.c (gimplify_save_expr): When in SSA form allow
17703         SAVE_EXPRs to compute to SSA vars.
17705 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
17707         PR target/85582
17708         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
17709         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
17710         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
17711         clobber operands[2], instead use a new pseudo.  Formatting fixes.
17713 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
17715         PR tree-optimization/85586
17716         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
17717         exit early for statements in the same group if the accesses are
17718         not strided.
17720 2018-05-02  Tom de Vries  <tom@codesourcery.com>
17722         PR lto/85451
17723         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
17724         error message.
17726 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
17728         PR tree-optimization/85143
17729         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
17731 2018-05-01  Tom de Vries  <tom@codesourcery.com>
17733         PR lto/85451
17734         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
17735         not found" error message.
17737 2018-05-01  Tom de Vries  <tom@codesourcery.com>
17739         PR other/83786
17740         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
17741         * vec.c (test_ordered_remove_if): New function.
17742         (vec_c_tests): Call test_ordered_remove_if.
17743         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
17744         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
17745         * tree-vect-patterns.c (vect_pattern_recog_1): Use
17746         VEC_ORDERED_REMOVE_IF.
17748 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
17750         PR tree-optimization/82665
17751         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
17752         pointer subtraction where arguments come from a memchr call.
17754 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
17756         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
17757         --push-state --as-needed and --pop-state instead of --as-needed and
17758         --no-as-needed if ld supports it.
17759         * configure: Regenerated.
17761         PR web/85578
17762         * doc/install.texi2html: Replace _002d with - and _002a with * in
17763         generated html files using sed.
17765 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17767         PR c++/85523
17768         * gcc-rich-location.c (blank_line_before_p): New function.
17769         (use_new_line): New function.
17770         (gcc_rich_location::add_fixit_insert_formatted): New function.
17771         * gcc-rich-location.h
17772         (gcc_rich_location::add_fixit_insert_formatted): New function.
17774 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17776         * selftest.c (assert_streq): Rename "expected" and "actual" to
17777         "val1" and "val2".  Extend NULL-handling to cover both inputs
17778         symmetrically, while still requiring both to be non-NULL for a pass.
17779         * selftest.h (assert_streq): Rename "expected" and "actual" to
17780         "val1" and "val2".
17781         (ASSERT_EQ): Likewise.
17782         (ASSERT_EQ_AT): Likewise.
17783         (ASSERT_KNOWN_EQ): Likewise.
17784         (ASSERT_KNOWN_EQ_AT): Likewise.
17785         (ASSERT_NE): Likewise.
17786         (ASSERT_MAYBE_NE): Likewise.
17787         (ASSERT_MAYBE_NE_AT): Likewise.
17788         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
17789         the assertion to pass.
17790         (ASSERT_STREQ_AT): Likewise.
17792 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
17794         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
17795         interaction with -pie.
17797 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17799         * selftest.h: Fix alphabetization of per-source-file selftest
17800         declarations.
17802 2018-04-30  Jason Merrill  <jason@redhat.com>
17804         PR c++/61982 - dead stores to destroyed objects.
17805         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
17806         of clobber.
17808 2018-04-30  Jason Merrill  <jason@redhat.com>
17810         * tree.c (build_clobber): New.
17811         * tree.h: Declare it.
17812         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
17814 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17816         * diagnostic-show-locus.c (layout::layout): Update for
17817         location_get_source_line returning a char_span.
17818         (struct char_span): Move to input.h.
17819         (struct correction): Update for fields in char_span becoming
17820         private.
17821         (struct source_line): Update for location_get_source_line
17822         returning a char_span.
17823         (layout::print_line): Likewise.
17824         * edit-context.c (edited_file::print_content): Likewise.
17825         (edited_file::print_diff_hunk): Likewise.
17826         (edited_file::print_run_of_changed_lines): Likewise.
17827         (edited_file::get_num_lines): Likewise.
17828         (edited_line::edited_line): Likewise.
17829         * final.c (asm_show_source): Likewise.
17830         * input.c (location_get_source_line): Convert return type
17831         from const char * to char_span, losing the final "line_len"
17832         param.
17833         (dump_location_info): Update for the above.
17834         (get_substring_ranges_for_loc): Likewise.  Use a char_span
17835         when handling the literal within the line.
17836         (test_reading_source_line): Update for location_get_source_line
17837         returning a char_span.
17838         * input.h (class char_span): Move here from
17839         diagnostic-show-locus.c, converting from a struct to a class.
17840         Make data members private.
17841         (char_span::operator bool): New.
17842         (char_span::length): New.
17843         (char_span::get_buffer): New.
17844         (char_span::operator[]): New.
17845         (char_span::subspan): Make const.
17846         (char_span::xstrdup): New.
17847         (location_get_source_line): Convert return type from const char *
17848         to char_span, losing the final "line_size" param.
17850 2018-04-30  Jan Hubicka  <jh@suse.cz>
17852         * lto-wrapper.c (ltrans_priorities): New static var.
17853         (cmp_priority): New.
17854         (run_gcc): Read priorities and if doing parallel build order
17855         the Makefile by them.
17857 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
17859         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
17861 2018-04-30  Richard Biener  <rguenther@suse.de>
17863         * tree-cfg.c (verify_address): Remove base argument, add
17864         flag whether to check TREE_ADDRESSABLE and do that.
17865         (verify_expr): Remove.
17866         (verify_types_in_gimple_reference): Add pieces from verify_expr.
17867         (verify_gimple_assign_single): Likewise.
17868         (verify_gimple_switch): Likewise.
17869         (verify_expr_location_1): Dereference tp once.  Add (disabled)
17870         piece from verify_expr.
17871         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
17873 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
17875         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
17877 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
17879         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
17880         (small_data_pattern): Likewise.
17881         (arc_rewrite_small_data): Likewise.
17882         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
17883         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
17884         (get_symbol_alignment): New function.
17885         (legitimate_small_data_address_p): Likewise.
17886         (legitimate_scaled_address): Update, call
17887         legitimate_small_data_address_p.
17888         (output_sdata): New static variable.
17889         (arc_print_operand): Update how we handle small data operands.
17890         (arc_print_operand_address): Likewise.
17891         (arc_legitimate_address_p): Update, use
17892         legitimate_small_data_address_p.
17893         (arc_rewrite_small_data_p): Remove.
17894         (arc_rewrite_small_data_1): Likewise.
17895         (arc_rewrite_small_data): Likewise.
17896         (small_data_pattern): Likewise.
17897         (compact_sda_memory_operand): Update to use
17898         legitimate_small_data_address_p and get_symbol_alignment.
17899         (prepare_move_operands): Don't rewite sdata pattern.
17900         (prepare_extend_operands): Remove.
17901         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
17902         pattern.
17903         (zero_extendqisi2): Likewise.
17904         (zero_extendhisi2): Likewise.
17905         (extendqihi2): Likewise.
17906         (extendqisi2): Likewise.
17907         (extendhisi2): Likewise.
17908         (addsi3): Likewise.
17909         (subsi3): Likewise.
17910         (andsi3): Likewise.
17911         * config/arc/constraints.md (Usd): Change it to memory constraint.
17913 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
17915         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
17916         as source of std instructions.
17917         * config/arc/arc.md (movsi_insn): Update pattern predicate to
17918         allow 6-bit constants as source for store instructions.
17919         (movdi_insn): Update instruction pattern to allow 6-bit constants
17920         as source for store instructions.
17922 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
17924         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
17926 2018-04-30  Nathan Sidwell  <nathan@acm.org>
17927             Sandra Loosemore <sandra@codesourcery.com>
17929         * dumpfile.c (dump_open): Allow '-' for stdout.
17930         * doc/invoke.texi (Developer Options): Document dump filename
17931         determination early.  Document stdin/stdout selection.
17933 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
17935         Microblaze Target: PIC data text relative
17937         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
17938         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
17939         Add declaration.
17940         * config/microblaze/microblaze.h (microblaze_constant_address_p):
17941         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
17942         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
17943         New addressing mode for data-text relative position indepenedent code.
17944         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
17945         'ADDRESS_SYMBOLIC_TXT_REL'.
17946         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
17947         (microblaze_legitimate_pic_operand): Exclude function calls from
17948         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
17949         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
17950         addresses cases.
17951         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
17952         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
17953         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
17954         for 'address + offset'.
17955         (microblaze_expand_prologue): Add new function prologue call for
17956         'r20' assignation.
17957         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
17958         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
17959         table in case of TARGET_PIC_DATA_TEXT_REL.
17960         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
17961         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
17962         Add new macros 'UNSPEC_TEXT',
17963         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
17964         + exclude function calls from 'UNSPEC_PLT' in case of data text
17965         relative mode.
17966         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
17967         new target hook for generating address diff vector tables in case of
17968         flag_pic.
17969         * doc/tm.texi : Regenerate.
17970         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
17971         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
17972         of addr diff vector generation.
17973         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
17974         target hook definition.
17975         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
17976         Add default function for generate_pic_addr_diff_vec -> flag_pic.
17977         * doc/invoke.texi (Add new pic option): Add new microblaze pic
17978         option for data text relative.
17980 2018-04-30  Richard Biener  <rguenther@suse.de>
17982         * tree-chrec.h (evolution_function_is_constant_p): Remove
17983         redundant check.
17984         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
17986 2018-04-30  Richard Biener  <rguenther@suse.de>
17988         PR bootstrap/85571
17989         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
17991 2018-04-30  Richard Biener  <rguenther@suse.de>
17993         PR tree-optimization/28364
17994         PR tree-optimization/85275
17995         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
17996         copying first exit test.
17998 2018-04-28  Mark Wielaard  <mark@klomp.org>
18000         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
18001         dwarf_version >= 5.
18002         (dwarf_AT): Handle DW_AT_addr_base.
18003         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
18005 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
18007         PR target/84431
18008         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
18009         (*ashl<dwi>3_doubleword_mask_1): Ditto.
18010         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
18011         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
18013 2018-04-28  Richard Biener  <rguenther@suse.de>
18015         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
18016         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
18017         to reflect use.  Only add interesting stmts.
18019 2018-04-27  Martin Jambor  <mjambor@suse.cz>
18021         PR ipa/85549
18022         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
18023         the jump function allows for passing through aggregate values.
18025 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
18027         * input.h (in_system_header_at): Convert from macro to inline
18028         function.
18029         (from_macro_expansion_at): Likewise.
18030         (from_macro_definition_at): Likewise.
18032 2018-04-27  Jeff Law  <law@redhat.com>
18034         * config.gcc: Mark tile* targets as deprecated/obsolete.
18036 2018-04-27  Richard Biener  <rguenther@suse.de>
18038         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
18039         fix for ILP32.
18041 2018-04-27  Richard Biener  <rguenther@suse.de>
18043         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
18045 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
18047         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
18048         with Yd constraint. Set "preferred_for_speed" attribute from
18049         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
18050         with Yd constraint.
18051         (*movdi_internal): Ditto.
18052         (movti_interunit splitters): Remove
18053         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
18054         (movdi_interunit splitters): Ditto.
18055         * config/i386/constraints.md (Ye): Remove.
18056         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
18058 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18060         PR target/85512
18061         * config/aarch64/constraints.md (Usg): Limit to 31.
18062         (Usj): Limit to 63.
18064 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
18066         PR tree-optimization/85529
18067         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
18068         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
18069         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
18070         zero extension or masking of the MSB bit.
18071         (optimize_range_tests): Add FIRST_BB argument, pass it through
18072         to optimize_range_tests_var_bound.
18073         (maybe_optimize_range_tests, reassociate_bb): Adjust
18074         optimize_range_tests callers.
18076 2018-04-26  Richard Biener  <rguenther@suse.de>
18077             Jakub Jelinek  <jakub@redhat.com>
18079         * cgraph.h (symbol_table): Just declare debug method here.
18080         * symtab.c (symbol_table::debug): Define.
18082 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
18084         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
18086 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
18088         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
18089         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
18090         (*movdi_internal): Substitute Yi and Yj constraint with x
18091         and Ym and Yn constraint with y constraint.  Update "isa"
18092         attribute and set "preferred_for_speed" attribute from
18093         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
18094         (*movsi_internal): Ditto.
18095         (*movdf_internal): Ditto.
18096         (*movsf_internal): Ditto.
18097         (*zero_extendsidi2): Ditto.
18098         * config/i386/sse.md (vec_set<mode>_0): Ditto.
18099         (sse2_loadld): Ditto.
18100         (*vec_extract<ssevecmodelower>_0): Ditto.
18101         (*vec_extractv4si_0_zext_sse4): Ditto.
18102         (vec_concatv2di): Ditto.
18103         (*vec_dup<mode>): Ditto.
18104         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
18105         * config/i386/constraints.md (Yi): Remove.
18106         (Yj): Remove.
18107         (Ym): Remove.
18108         (Yn): Remove.
18110 2018-04-26  Nathan Sidwell  <nathan@acm.org>
18112         * dumpfile.c (dump_open): New.
18113         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
18114         (dump_finish): Detect stdio/stderr by value not name.
18116 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
18118         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
18120 2018-04-26  Tom de Vries  <tom@codesourcery.com>
18122         PR target/84952
18123         * config/nvptx/nvptx.c (verify_neutering_jumps)
18124         (verify_neutering_labels): New function
18125         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
18127 2018-04-26  Tom de Vries  <tom@codesourcery.com>
18129         PR target/84025
18130         * config/nvptx/nvptx.c (needs_neutering_p): New function.
18131         (nvptx_single): Use needs_neutering_p to skip over insns that do not
18132         need neutering.
18134 2018-04-26  Richard Biener <rguenther@suse.de>
18135             Tom de Vries  <tom@codesourcery.com>
18137         PR lto/85422
18138         * lto-streamer-out.c (output_function): Fixup loops if required to match
18139         discovery done in the reader.
18141 2018-04-26  Richard Biener  <rguenther@suse.de>
18143         PR tree-optimization/85116
18144         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
18145         have a loop exit from the single latch predecessor.  Remove
18146         case of header with just condition.
18147         (ch_base::copy_headers): Exclude infinite loops from any
18148         processing.
18149         (pass_ch::execute): Record exits.
18151 2018-04-26  Richard Biener  <rguenther@suse.de>
18153         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
18154         prologue cost vector and pass it to vect_get_load_cost.
18155         (vect_get_peeling_costs_all_drs): Likewise.
18156         (vect_peeling_hash_get_lowest_cost): Likewise.
18157         (vect_enhance_data_refs_alignment): Likewise.
18159 2018-04-26  Richard Biener  <rguenther@suse.de>
18161         PR middle-end/85450
18162         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
18163         checking of integer<->pointer conversions.
18164         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
18165         sign-/zero-extending pointer types.
18166         (expand_omp_for_static_chunk): Likewise.
18168 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
18169             Jean Lee  <xiaoyur347@gmail.com>
18171         * config/mips/mips.c (mips_asan_shadow_offset): New function.
18172         (TARGET_ASAN_SHADOW_OFFSET): Define.
18173         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
18174         true for -fsanitize=address.
18176 2018-04-25  Mark Wielaard  <mark@klomp.org>
18178         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
18179         shorter ones.
18181 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
18183         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
18184         than "alu", remove explicit "memory" and "imm_disp" attributes.
18185         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
18187         PR middle-end/85414
18188         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
18189         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
18190         gen_lowpart_no_emit.
18192 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
18194         PR target/85473
18195         * config/i386/i386.c (ix86_expand_builtin): Change memory
18196         operand to XI, extend p0 to Pmode.
18197         * config/i386/i386.md: Change unspec volatile and operand
18198         1 mode to XI, change operand 0 mode to P.
18200 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18202         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
18203         GET_MODE_MASK before any checking.
18204         (nds32_can_use_bset_p): Likewise.
18205         (nds32_can_use_btgl_p): Likewise.
18207 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18209         * config/nds32/nds32-doubleword.md: New define_split pattern for
18210         illegal register number.
18212 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18214         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
18216 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
18218         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
18220 2018-04-25  Richard Biener  <rguenther@suse.de>
18222         * lto-streamer.h (LTO_major_version): Bump to 8.
18224 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
18226         * BASE-VER: Set to 9.0.0.
18228 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
18230         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
18231         in __abskf2 and __powikf2.
18233 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18235         PR target/85512
18236         * config/aarch64/constraints.md (Usg, Usj): New constraints.
18237         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
18238         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
18239         Use the above on operand 2.  Reindent.
18240         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
18242 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
18244         PR target/85485
18245         * common/config/i386/i386-common.c (ix86_handle_option): Don't
18246         handle OPT_mcet.
18247         * config/i386/i386.opt (mcet): Removed.
18248         * doc/install.texi: Remove -mcet documentation.
18249         * doc/invoke.texi: Likewise.
18251 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
18253         PR target/85485
18254         * doc/install.texi: Remove -mcet from bootstrap-cet.
18256 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
18258         PR target/85511
18259         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
18260         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
18261         if TARGET_64BIT.
18263         PR target/85503
18264         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
18265         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
18266         containing a CONST_VECTOR.
18268 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
18270         * doc/install.texi: Update newlib dependency for nvptx.
18272 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
18274         PR target/85508
18275         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
18276         instead of INTVAL when shifting x left.
18278 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
18280         PR tree-optimization/85478
18281         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
18282         vect_grouped_store_supported for single element vectors.
18284 2018-04-24  Richard Biener  <rguenther@suse.de>
18286         PR target/85491
18287         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
18288         load cost increase to the case of non-constant step.
18290 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
18292         PR target/84828
18293         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
18294         destination if any_malformed_asm.
18296 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
18298         PR middle-end/85496
18299         * expr.c (store_field): In the bitfield case, if the value comes from
18300         a function call and is returned in registers by means of a PARALLEL,
18301         do not change the mode of the temporary unless BLKmode and VOIDmode.
18303 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
18305         PR rtl-optimization/85423
18306         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
18307         dependencies to debug insns when the previous insn is non-debug.
18309 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
18311         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
18312         enums into a single definition.
18313         (fls): Fix predicates and printing.
18314         (seti): Likewise.
18316 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
18318         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
18319         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
18320         and short u6 immediate.
18321         (check_if_valid_sleep_operand): Remove.
18322         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
18324 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18326         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
18327         flag_always_save_lp condition.
18328         * config/nds32/nds32.opt (malways-save-lp): New option.
18330 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
18332         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
18333         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
18334         * config/nds32/nds32.h
18335         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
18336         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
18338 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
18340         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
18341         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
18343 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
18344             Chung-Ju Wu  <jasonwucj@gmail.com>
18346         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
18347         Declare.
18348         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
18349         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
18351 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18353         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
18355 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18357         * config/nds32/nds32-protos.h (nds32_data_alignment,
18358         nds32_local_alignment): Declare.
18359         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
18360         nds32_local_alignment): New functions.
18361         (TARGET_CONSTANT_ALIGNMENT): Define.
18362         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
18364 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18366         * config/nds32/nds32.c
18367         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
18368         (TARGET_MODES_TIEABLE_P): Likewise.
18370 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
18372         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
18373         level Ofast and Og.
18375 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
18376             Chung-Ju Wu  <jasonwucj@gmail.com>
18378         * config/nds32/constants.md (unspec_volatile_element): Add enum values
18379         for unaligned access.
18380         * config/nds32/nds32-intrinsic.c: Implementation of expanding
18381         unaligned access.
18382         * config/nds32/nds32-intrinsic.md: Likewise.
18383         * config/nds32/nds32_intrinsic.h: Likewise.
18384         * config/nds32/nds32.h (nds32_builtins): Likewise.
18385         * config/nds32/nds32.opt (munaligned-access): New option.
18386         * config/nds32/nds32.c (nds32_asm_file_start): Display
18387         flag_unaligned_access status.
18389 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
18391         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
18392         -mno-relax is present.
18393         * config/riscv/linux.h (LINK_SPEC): Ditto.
18395 2018-04-20  Martin Sebor  <msebor@redhat.com>
18397         PR c/85365
18398         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
18399         for null pointers.
18400         (gimple_fold_builtin_stxcpy_chk): Same.
18401         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
18403 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
18405         PR target/85456
18406         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
18407         __powikf2 when long double is IEEE 128-bit.
18409 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
18411         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
18412         step to make sure stack always aligned.
18414 2018-04-20  Carl Love  <cel@us.ibm.com>
18416         PR target/83402
18417         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
18418         size check for arg0.
18420 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
18421             Tom de Vries  <tom@codesourcery.com>
18423         PR target/85445
18424         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
18425         Emit insns for calls too.
18426         (nvptx_find_par): Always look for worker-level predecessor insn.
18427         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
18428         calls.
18429         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
18430         (nvptx_process_pars): Propagate frames for calls.
18432 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
18434         PR target/85469
18435         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
18436         Removed.
18437         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
18438         (ix86_handle_option): Don't handle OPT_mibt.
18439         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
18440         __SHSTK__.
18441         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
18442         has_ibt and ibt.
18443         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
18444         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
18445         (ix86_target_macros): Define __CET__ with flag_cf_protection
18446         for -fcf-protection.
18447         * config/i386/i386.c (isa2_opts): Remove -mibt.
18448         * config/i386/i386.h (TARGET_IBT): Removed.
18449         (TARGET_IBT_P): Likewise.
18450         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
18451         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
18452         * config/i386/i386.opt (mcet): Update help message.
18453         (mshstk): Likewise.
18454         (mibt): Removed.
18455         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
18456         -mcet as an alias for -mshstk.
18458 2018-04-20  Richard Biener <rguenther@suse.de>
18460         PR middle-end/85475
18461         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
18462         complexity by forcing a single use of the multiply operand.
18464 2018-04-20  Martin Jambor  <mjambor@suse.cz>
18466         ipa/85449
18467         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
18468         recursion dependency to only apply to non-clones.
18470 2018-04-20  Martin Jambor  <mjambor@suse.cz>
18472         ipa/85447
18473         * ipa-cp.c (create_specialized_node): Check that clones of
18474         self-recursive edges exist during IPA-CP.
18476 2018-04-19  Toon Moene  <toon@moene.org>
18478         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
18479         by -O3.
18481 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
18483         PR tree-optimization/85467
18484         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
18485         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
18486         VECTOR_CST element to type.
18488 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
18490         PR target/85397
18491         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
18492         * config/i386/i386.md (builtin_setjmp_setup): Removed.
18493         (builtin_longjmp): Likewise.
18494         (save_stack_nonlocal): New pattern.
18495         (restore_stack_nonlocal): Likewise.
18497 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
18499         PR target/85404
18500         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
18501         Replace ASM_OUTPUT_LABEL with fprintf.
18503 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
18505         PR target/85417
18506         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
18507         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
18508         * config/i386/i386-c.c (ix86_target_macros_internal): Also
18509         define __IBT__ and __SHSTK__ for -fcf-protection.
18510         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
18511         TARGET_IBT.
18512         (ix86_trampoline_init): Likewise.
18513         (x86_output_mi_thunk): Likewise.
18514         (ix86_notrack_prefixed_insn_p): Likewise.
18515         (ix86_option_override_internal): Don't disallow -fcf-protection.
18516         * config/i386/i386.md (rdssp<mode>): Also enable for
18517         -fcf-protection.
18518         (incssp<mode>): Likewise.
18519         (nop_endbr): Likewise.
18520         * config/i386/i386.opt (mcet): Change help message to built-in
18521         functions only.
18522         (mibt): Likewise.
18523         (mshstk): Likewise.
18524         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
18525         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
18526         enable CET built-in functions.
18528 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
18530         * common/config/i386/i386-common.c
18531         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
18532         OPTION_MASK_ISA_MOVDIRI_UNSET,
18533         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
18534         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
18535         * config.gcc (movdirintrin.h): New header.
18536         * config/i386/cpuid.h (bit_MOVDIRI,
18537         bit_MOVDIR64B): New bits.
18538         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
18539         and -mmvodir64b.
18540         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
18541         (VOID, PVOID, PCVOID)): New function types.
18542         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
18543         __builtin_ia32_directstoreu_u64,
18544         __builtin_ia32_movdir64b): New builtins.
18545         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
18546         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
18547         and -mmovdiri.
18548         (ix86_valid_target_attribute_inner_p): Ditto.
18549         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
18550         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
18551         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
18552         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
18553         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
18554         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
18555         (movdiri<mode>, movdir64b_<mode>): New.
18556         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
18557         * config/i386/immintrin.h: Include movdirintrin.h.
18558         * config/i386/movdirintrin.h: New file.
18559         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
18561 2018-04-19  Richard Biener  <rguenther@suse.de>
18563         PR middle-end/85455
18564         * cfg.c (clear_bb_flags): When loop state says we have
18565         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
18567 2018-04-19  Richard Biener  <rguenther@suse.de>
18569         PR tree-optimization/84737
18570         * tree-vect-data-refs.c (vect_copy_ref_info): New function
18571         copying restrict info.
18572         (vect_setup_realignment): Use it.
18573         * tree-vectorizer.h (vect_copy_ref_info): Declare.
18574         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
18575         the first DR to all generated stores.
18576         (vectorizable_load): Likewise for loads.
18578 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
18580         PR tree-optimization/85446
18581         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
18582         the integral and pointer types to have the same precision.
18584         * doc/install.texi: Document --disable-cet being the default and
18585         --enable-cet=auto.
18587 2018-04-18  Martin Liska  <mliska@suse.cz>
18589         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
18590         style.
18592 2018-04-18  Martin Liska  <mliska@suse.cz>
18594         Revert
18595         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
18597         PR ipa/83983
18598         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
18599         arguments if they are comparable.
18601 2018-04-18  Martin Liska  <mliska@suse.cz>
18603         Revert
18604         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
18606         PR lto/84805
18607         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
18608         incomplete types.
18610 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
18612         PR target/85388
18613         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
18614         ENDBR after calling __morestack.
18616 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
18618         PR jit/85384
18619         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
18620         by using gcc_base_ver to generate a gcc_driver_version, and use
18621         it when generating GCC_DRIVER_NAME.
18622         * configure: Regenerate.
18624 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
18626         PR target/81084
18627         * config.gcc: Obsolete powerpc*-*-*spe*.
18629 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18631         PR debug/84637
18632         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
18633         (stabstr_D): Change type of unum from unsigned int to
18634         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
18635         type.
18637 2018-04-17  Jim Wilson  <jimw@sifive.com>
18639         PR 84856
18640         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
18641         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
18642         Set arg_pointer_offset after using pretend_args_size.
18644 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18646         PR rtl-optimization/85431
18647         * dse.c (record_store): Ignore zero width stores.
18649         PR sanitizer/85230
18650         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
18651         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
18652         __builtin_stack_restore rather than after it.
18653         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
18654         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
18655         argument instead of virtual_dynamic_stack_rtx.
18657 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18659         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
18660         New prototype.
18661         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
18662         Add note to error message to explain internal mapping of overloaded
18663         built-in function name to non-overloaded built-in function name.
18664         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
18665         function.
18667 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
18669         PR target/85424
18670         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
18671         where the inputs overlap with the output.
18673 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18675         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
18676         (=v, v) alternative and explicit "memory" attribute.
18677         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
18678         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18679         attributes.
18680         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
18681         "sselog1" type instead of "sselog".
18682         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
18683         "sselog".  Remove explicit "memory" attribute.
18684         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
18685         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18686         attributes.
18687         (vec_extract_hi_v32hi): Merge all alternatives into one, use
18688         "sselog1" type instead of "sselog".  Remove explicit "memory"
18689         attribute.
18690         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
18691         use "sselog1" type instead of "sselog".  Remove explicit "memory"
18692         attribute.
18693         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
18694         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
18695         attributes.
18696         (vec_extract_hi_v64qi): Merge all alternatives into one, use
18697         "sselog1" type instead of "sselog".  Remove explicit "memory"
18698         attribute.
18699         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
18700         use "sselog1" type instead of "sselog".  Remove explicit "memory"
18701         attribute.
18703         PR target/85430
18704         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
18706         PR middle-end/85414
18707         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
18708         on a SUBREG.
18710 2018-04-17  Martin Jambor  <mjambor@suse.cz>
18712         PR ipa/85421
18713         * ipa-cp.c (create_specialized_node): Call
18714         expand_all_artificial_thunks if necessary.
18716 2018-04-17  Martin Liska  <mliska@suse.cz>
18718         PR lto/85405
18719         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
18720         in message, remote space in between '_G' and '('.
18722 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
18724         PR target/85281
18725         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
18726         avx512f_vmcmp<mode>3<round_saeonly_name>,
18727         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
18728         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
18729         avx512f_rndscale<mode><round_saeonly_name>,
18730         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
18731         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
18732         Use %<iptr>2 instead of %2 for -masm=intel.
18733         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
18734         avx512f_vcvttss2usi<round_saeonly_name>,
18735         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
18736         -masm=intel.
18737         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
18738         avx512f_vcvttsd2usi<round_saeonly_name>,
18739         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
18740         Use %q1 instead of %1 for -masm=intel.
18741         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
18742         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
18743         of %3 for -masm=intel.
18744         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
18745         -masm=intel.
18746         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
18747         -masm=intel.
18748         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
18749         -masm=intel.
18750         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
18751         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
18752         %g1.
18753         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
18754         -masm=intel.
18755         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
18756         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
18757         %g1 and one with %0 and %1.
18758         (avx512er_vmrcp28<mode><round_saeonly_name>,
18759         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
18760         %1 for -masm=intel.
18761         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
18762         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
18763         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
18764         of %0 and %{%4%} for -masm=intel.
18765         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
18766         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
18767         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
18768         order of %0 and %{%5%}%{z%} for -masm=intel.
18770 2018-04-17  Jan Hubicka  <jh@suse.cz>
18772         PR lto/85405
18773         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
18775 2018-04-17  Martin Liska  <mliska@suse.cz>
18777         PR ipa/85329
18778         * multiple_target.c (create_dispatcher_calls): Set apostrophes
18779         for target_clone error message.  Make default implementation
18780         clone to be a local declaration.
18781         (separate_attrs): Add new argument and check for an empty
18782         string.
18783         (expand_target_clones): Handle it.
18784         (ipa_target_clone): Make redirection just for target_clones
18785         functions.
18787 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
18788             Tom de Vries  <tom@codesourcery.com>
18790         PR middle-end/84955
18791         * omp-expand.c (expand_oacc_for): Add dummy false branch for
18792         tiled basic blocks without omp continue statements.
18794 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
18796         PR target/83660
18797         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
18798         vec_extract expression as having side effects to make sure it gets
18799         a cleanup point.
18801 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
18803         PR target/85403
18804         * config/i386/i386.c (get_builtin_code_for_version): Check
18805         error_mark_node.
18807 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
18809         PR target/84331
18810         * config.gcc: Support "skylake".
18811         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18812         PROCESSOR_SKYLAKE.
18813         * config/i386/i386.c (m_SKYLAKE): Define.
18814         (processor_target_table): Add "skylake".
18815         (ix86_option_override_internal): Add "skylake".
18816         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
18817         PROCESSOR_CANNONLAKE.
18818         (get_builtin_code_for_version): Fix priority for
18819         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
18820         PROCESSOR_SKYLAKE-AVX512.
18821         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
18822         (processor_type): Add PROCESSOR_SKYLAKE.
18824 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
18825             Jason Merrill  <jason@redhat.com>
18827         PR c++/85112
18828         * convert.c (convert_to_integer_1): Use direct recursion for
18829         enumeral types and types with a precision less than the number
18830         of bits in their mode.
18832 2018-04-16  Julia Koval  <julia.koval@intel.com>
18834         PR target/84413
18835         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
18836         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
18838 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
18840         PR target/85293
18841         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
18842         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
18843         and -mno-direct-move.
18845 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
18847         PR target/83402
18848         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
18849         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
18850         Ensure negative shifts result in {0}.
18852 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
18854         PR rtl-optimization/79916
18855         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
18856         regs (if any) to define how to gnerate SD moves when LRA is in
18857         progress.
18859 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
18861         PR rtl-optimization/85393
18862         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
18863         * except.c (expand_dw2_landing_pad_for_region): Make static.
18864         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
18865         a label and unconditional jump to old_bb, rather than
18866         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
18867         basic block.
18869         PR rtl-optimization/85376
18870         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
18871         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
18872         instead of a specific value.
18874 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
18875             Bin Cheng  <bin.cheng@arm.com>
18877         PR tree-optimization/82965
18878         PR tree-optimization/83991
18879         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
18880         by_profile_only parameter.
18881         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
18882         information if the loop was predicted to iterate too many times.
18883         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
18885 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
18887         PR lto/71991
18888         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
18889         always inline.
18891 2018-04-13  Martin Liska  <mliska@suse.cz>
18892             Jakub Jelinek  <jakub@redhat.com>
18894         PR middle-end/81657
18895         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
18896         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
18897         * builtins.c (expand_builtin_memory_copy_args): Use
18898         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
18899         handle dest_addr == pc_rtx.
18901 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
18903         PR target/85291
18904         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
18905         asked to not generate direct moves.
18906         (fix_trunc<mode>si2_stfiwx): Similar.
18907         (fix_trunc<mode>si2_internal): Similar.
18909 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
18911         PR debug/83157
18912         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
18913         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
18914         lookup if dest in some wider mode is known to be const0_rtx and
18915         if so, record permanent equivalence for it to be ZERO_EXTEND of
18916         the narrower mode destination.
18918 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
18920         * lto-streamer-out.c (output_function): Revert 259346.
18921         * omp-expand.c (expand_oacc_for): Likewise.
18923 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
18925         PR rtl-optimization/85354
18926         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
18927         * sel-sched.c (sel_global_init): ... here.
18929 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
18931         PR target/85238
18932         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
18933         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
18934         mode for PE-COFF targets.
18935         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
18936         (i386_pe_asm_lto_end): Likewise.
18937         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
18938         (TARGET_ASM_LTO_END): Likewise.
18939         * config/i386/winnt.c (saved_debug_info_level): New static variable.
18940         (i386_pe_asm_lto_start): New function.
18941         (i386_pe_asm_lto_end): Likewise.
18943 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
18944             Richard Biener  <rguenther@suse.de>
18946         PR middle-end/84955
18947         * lto-streamer-out.c (output_function): Fix CFG loop state before
18948         streaming out.
18949         * omp-expand.c (expand_oacc_for): Handle calls to internal
18950         functions like regular functions.
18952 2018-04-12  Richard Biener  <rguenther@suse.de>
18954         PR lto/85371
18955         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
18956         for the early LTO debug to properly generate references to it
18957         during DIE emission.  Do not re-use that for the skeleton for
18958         split-dwarf.
18959         (dwarf2out_early_finish): Likewise.
18961 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
18963         PR target/85328
18964         * config/i386/sse.md
18965         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
18966         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
18967         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
18968         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
18969         and output is a reg, avoid creating invalid lowpart subreg, but
18970         instead split into a 512-bit move.  Don't split if not AVX512VL,
18971         input is xmm16+ reg and output is a mem.
18972         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
18973         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
18974         xmm16+ reg and output is a mem.
18976 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18978         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
18979         also for flag_dwarf2_cfi_asm.
18981 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
18983         PR rtl-optimization/85342
18984         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
18985         a bool scalar var inside of the loop instead.  Don't try to update
18986         recog_data.operand after failed apply_change_group.
18988 2018-04-12  Tom de Vries  <tom@codesourcery.com>
18990         PR target/85296
18991         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
18992         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
18993         array with flexible array member as array without given dimension.
18994         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
18995         argument for undefined param to true.
18997 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
18999         PR target/85321
19000         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19001         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
19002         from PowerPC section.
19003         * config/rs6000/sysv4.opt (mcall-): Improve help text.
19004         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
19005         help text that is too long.
19006         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
19007         help text that is too long.
19008         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
19009         help text that is too long.
19011 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
19013         * config/alpha/alpha.md (stack_probe_internal): Rename
19014         from "probe_stack".  Update all callers.
19016 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19018         PR rtl-optimization/84566
19019         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
19020         sched_macro_fuse_insns.
19022 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19024         PR target/84301
19025         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
19026         (compute_block_dependences): ... from here.
19028 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19030         PR tree-optimization/85331
19031         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
19032         from int to HOST_WIDE_INT.
19034 2018-04-11  Martin Jambor  <mjambor@suse.cz>
19036         PR ipa/84149
19037         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
19038         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
19039         not the same as the source val.
19040         (cgraph_edge_brings_value_p): New parameter.
19041         (gather_edges_for_value): Pass destination value to
19042         cgraph_edge_brings_value_p.
19043         (perhaps_add_new_callers): Likewise.
19044         (get_info_about_necessary_edges): Likewise and exclude values brought
19045         only by self-recursive edges.
19046         (create_specialized_node): Redirect only clones of self-calling edges.
19047         (+self_recursive_pass_through_p): New function.
19048         (find_more_scalar_values_for_callers_subset): Use it.
19049         (find_aggregate_values_for_callers_subset): Likewise.
19050         (known_aggs_to_agg_replacement_list): Removed.
19051         (decide_whether_version_node): Re-calculate known constants for all
19052         remaining context clones.
19054 2018-04-11  Richard Biener  <rguenther@suse.de>
19056         PR lto/85339
19057         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
19058         from early DWARF output.
19059         (dwarf2out_early_finish): Output line info unconditionally into
19060         early DWARF and add reference to it.
19062 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19064         PR target/85281
19065         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
19066         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
19067         other than V2DFmode using iptr mode attribute.
19068         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
19070 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
19072         PR rtl-optimization/84659
19073         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
19075 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
19077         PR debug/85302
19078         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
19079         SIZEP is NULL.
19080         (output_loc_list): Pass address of a dummy size variable even in the
19081         locview handling loop.
19082         (index_location_lists): Add comment on why skip_loc_list_entry can't
19083         call size_of_locs.
19085 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
19087         PR target/85261
19088         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
19089         into register.
19091 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
19093         PR target/85321
19094         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
19095         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
19096         and -mstring-compare-inline-limit.
19098 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19100         PR target/85287
19101         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
19102         for stack clash protection in a register whenever we need it to be in
19103         a register.
19105 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19107         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
19108         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
19110 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
19112         PR target/85321
19113         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
19114         the help text.
19115         (mlong-double-): Ditto.
19116         * config/rs6000/sysv4.opt (msdata=): Ditto.
19117         (mtls-size=): Ditto.
19119 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19121         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
19122         erroneous entries for
19123         "vector int vec_ldl (int, long int *)", and
19124         "vector unsigned int vec_ldl (int, unsigned long int *)".
19125         Add comments and entries for
19126         "vector bool char vec_ldl (int, bool char *)",
19127         "vector bool short vec_ldl (int, bool short *)",
19128         "vector bool int vec_ldl (int, bool int *)",
19129         "vector bool long long vec_ldl (int, bool long long *)",
19130         "vector pixel vec_ldl (int, pixel *)",
19131         "vector long long vec_ldl (int, long long *)",
19132         "vector unsigned long long vec_ldl (int, unsigned long long *)".
19133         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
19134         type tree bool_long_long_type_node and correct definition of
19135         bool_V2DI_type_node to make reference to this new type tree.
19136         (rs6000_mangle_type): Replace erroneous reference to
19137         bool_long_type_node with bool_long_long_type_node.
19138         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
19139         comments to emphasize sign distinctions for char and int types and
19140         replace RS6000_BTI_bool_long constant with
19141         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
19142         use of RS6000_BTI_pixel.
19143         (bool_long_type_node): Remove this macro definition.
19144         (bool_long_long_type_node): New macro definition
19146 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
19148         PR rtl-optimization/85300
19149         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
19150         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
19151         simplify_unary_operation fails.
19153 2018-04-10  Martin Liska  <mliska@suse.cz>
19155         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
19156         cgraph_edge and ipa_ref.
19158 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
19160         PR target/85177
19161         PR target/85255
19162         * config/i386/sse.md
19163         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
19164         computation of the VEC_MERGE selector from mask.
19165         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
19166         Fix decoding of the VEC_MERGE selector into mask.
19168 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
19170         PR tree-optimization/85286
19171         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
19173 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
19175         * final.c (final_1): Set insn_last_address as well as
19176         insn_current_address.
19178 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19180         PR target/85173
19181         * explow.c (emit_stack_probe): Call validize_mem on memory location
19182         before passing it to gen_probe_stack.  Create address operand and
19183         legitimize it for the probe_stack_address case.
19185 2018-04-09  Jan Hubicka  <jh@suse.cz>
19187         PR lto/85078
19188         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
19189         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
19190         * tree.c (free_lang_data_in_type): Fix handling of binfos;
19191         walk basetypes.
19192         (free_lang_data): Rebuild type inheritance graph.
19194 2018-04-09  Martin Sebor  <msebor@redhat.com>
19196         * invoke.texi (-finline-small-functions): Mention other optimization
19197         options.
19198         (-findirect-inlining, -fpartial-inlining): Same.
19199         (-finline-functions-called-once): Same.
19200         (-freorder-blocks-and-partition): Same.
19202 2018-04-09  Jan Hubicka  <jh@suse.cz>
19204         PR rtl/84058
19205         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
19206         jumps; choose last target that matches the criteria (i.e.
19207         no partition changes for non-crossing jumps).
19208         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
19209         support for redirecting crossing jumps to non-crossing.
19211 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
19213         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
19214         also for naked functions.
19216 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
19218         * config/arc/arc.md (add_shift): New pattern.
19219         (add_shift2): Likewise.
19220         (sub_shift): Likewise.
19221         (sub_shift_cmp0_noout): Likewise.
19222         (compare_si_ashiftsi): Likewise.
19223         (xbfu_cmp0_noout): New combine pattern.
19224         (xbfu_cmp0"): Likewise.
19225         (movsi_set_cc_insn): Place the predicable variant first.
19226         (commutative_binary_cmp0_noout): Remove clobber.
19227         (commutative_binary_cmp0): New pattern.
19228         (noncommutative_binary_cmp0): Likewise.
19229         (noncommutative_binary_cmp0_noout): Likewise.
19230         (noncommutative_binary_comparison_result_used): Removed.
19231         (rsub_cmp0): New pattern.
19232         (rsub_cmp0_noout): Likewise.
19233         (extzvsi): Changed, keep only meaningful variants.
19234         (SQH, SEZ): New iterators.
19235         (SQH_postfix): New mode attribute.
19236         (SEZ_prefix): New code attribute.
19237         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
19238         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
19239         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
19240         of numerical value.
19241         (noncommutative_operator): Check the availability of barrel
19242         shifter option.
19244 2018-04-09  Richard Biener  <rguenther@suse.de>
19246         PR tree-optimization/85284
19247         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
19248         Only use the niter constraining form of simple_iv when the exit
19249         is always executed.
19251 2018-04-09  Tom de Vries  <tom@codesourcery.com>
19253         PR target/84041
19254         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
19255         (define_expand "*memory_barrier"): New define_expand.
19256         (define_insn "memory_barrier"): New insn.
19258 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19260         PR rtl-optimization/80463
19261         PR rtl-optimization/83972
19262         PR rtl-optimization/83480
19264         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
19265         correct producer for the insn.
19266         (tidy_control_flow): Fixup seqnos in case of debug insns.
19268 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19270         PR rtl-optimization/83913
19272         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
19273         different sched-times when merging exprs.
19275 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19277         PR rtl-optimization/83962
19279         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
19280         tidy_fallthru_edge and tidy_control_flow.
19282 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
19284         PR rtl-optimization/83530
19286         * sel-sched.c (force_next_insn): New global variable.
19287         (remove_insn_for_debug): When force_next_insn is true, also leave only
19288         next insn in the ready list.
19289         (sel_sched_region): When the region wasn't scheduled, make another pass
19290         over it with force_next_insn set to 1.
19292 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
19294         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
19295         into tm_file.
19296         * config/nds32/constants.md (unspec_volatile_element): Add enum values
19297         for interrupt control.
19298         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
19299         functions for interrupt control.
19300         * config/nds32/nds32-intrinsic.md: Likewise.
19301         * config/nds32/nds32_intrinsic.h: Likewise.
19302         * config/nds32/nds32.h (nds32_builtins): Likewise.
19304 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
19306         * config/nds32/nds32.c (nds32_init_machine_status,
19307         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
19308         strict_aligned_p field.
19309         (nds32_expand_to_rtl_hook): New function.
19310         (TARGET_EXPAND_TO_RTL_HOOK): Define.
19311         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
19313 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19314             Chung-Ju Wu  <jasonwucj@gmail.com>
19316         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
19317         * config/nds32/nds32-n7.md: New file.
19318         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
19319         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
19320         pipeline.
19321         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
19322         * config/nds32/nds32.md (pipeline_model): Add n7.
19323         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
19324         * config/nds32/pipelines.md: Include n7 settings.
19326 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19327             Chung-Ju Wu  <jasonwucj@gmail.com>
19329         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
19330         * config/nds32/nds32-e8.md: New file.
19331         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
19332         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
19333         pipeline.
19334         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
19335         * config/nds32/nds32.md (pipeline_model): Add e8.
19336         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
19337         * config/nds32/pipelines.md: Include e8 settings.
19339 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19340             Chung-Ju Wu  <jasonwucj@gmail.com>
19342         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
19343         * config/nds32/nds32-n8.md: New file.
19344         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
19345         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
19346         pipeline.
19347         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
19348         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
19349         * config/nds32/nds32.md (pipeline_model): Add n8.
19350         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
19351         * config/nds32/pipelines.md: Include n8 settings.
19353 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
19354             Chung-Ju Wu  <jasonwucj@gmail.com>
19356         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
19357         * config/nds32/nds32-n9-2r1w.md: New file.
19358         * config/nds32/nds32-n9-3r2w.md: New file.
19359         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
19360         nds32_register_ports): New or modify for cpu n9.
19361         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
19362         pipeline.
19363         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
19364         * config/nds32/nds32-utils.c: New file.
19365         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
19366         TARGET_MUL_SLOW): Define.
19367         * config/nds32/nds32.md (pipeline_model): New attribute.
19368         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
19369         New options that support cpu n9.
19370         * config/nds32/pipelines.md: Include n9 settings.
19371         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
19373 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
19375         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
19376         information if necessary.
19377         (output_cond_branch_compare_zero): Likewise.
19378         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
19379         (nds32_target_alignment): Refine for alignment.
19380         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
19381         (FUNCTION_BOUNDARY): Modify.
19382         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
19383         align case.
19384         * config/nds32/nds32.opt (malways-align, malign-functions): New.
19386 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
19388         * config/nds32/constants.md (unspec_volatile_element): Add values for
19389         TLB operation and data prefetch.
19390         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
19391         functions for TLB operation and data prefetch.
19392         * config/nds32/nds32-intrinsic.md: Likewise.
19393         * config/nds32/nds32_intrinsic.h: Likewise.
19394         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
19395         (nds32_print_operand): Likewise.
19396         * config/nds32/nds32.h (nds32_builtins): Likewise.
19398 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
19399         Andrew Pinski <pinsika@gcc.gnu.org>
19401         PR middle-end/82976
19402         * match.pd: Use constant_boolean_node of correct type instead of
19403         boolean_true_node or boolean_false_node for simplifying
19404         pointer comparisons to zero.
19406 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
19408         PR tree-optimization/80021
19409         * tree.c (verify_type_variant): Make error call in verify_variant_match
19410         translatable and remove final full stop.
19412 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
19414         * config/nds32/constants.md (unspec_volatile_element): Add
19415         UNSPEC_VOLATILE_EH_RETURN.
19416         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
19417         nds32_output_stack_pop): Support dwarf exception handling process.
19418         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
19419         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
19420         exception handling process.
19421         (nds32_compute_stack_frame): Likewise.
19422         (nds32_return_addr_rtx): Likewise.
19423         (nds32_initial_elimination_offset): Likewise.
19424         (nds32_expand_prologue): Likewise.
19425         (nds32_expand_epilogue): Likewise.
19426         (nds32_dynamic_chain_address): New function.
19427         * config/nds32/nds32.h (machine_function): Add fields for dwarf
19428         exception handling.
19429         (DYNAMIC_CHAIN_ADDRESS): Define.
19430         (EH_RETURN_DATA_REGNO): Define.
19431         (EH_RETURN_STACKADJ_RTX): Define.
19432         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
19433         patterns for dwarf exception handling.
19435 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
19437         * config/nds32/nds32.h: Clean up obsolete macros.
19439 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19441         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19442         Add enum values for particular instructions.
19443         * config/nds32/nds32-intrinsic.c: Implementation of expanding
19444         particular intrinsic functions.
19445         * config/nds32/nds32-intrinsic.md: Likewise.
19446         * config/nds32/nds32_intrinsic.h: Likewise.
19447         * config/nds32/nds32.h (nds32_builtins): Likewise.
19448         * config/nds32/nds32.md (type): Add pbsad and pbsada.
19449         (btst, ave): New patterns for particular instructions.
19451 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19453         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19454         Add enum values for atomic load/store and memory sync.
19455         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
19456         and memory sync.
19457         * config/nds32/nds32-intrinsic.md: Likewise.
19458         * config/nds32/nds32_intrinsic.h: Likewise.
19459         * config/nds32/nds32.h (nds32_builtins): Likewise.
19461 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
19463         PR tree-optimization/85257
19464         * fold-const.c (native_encode_vector): If not all elts could fit
19465         and off is -1, return 0 rather than offset.
19466         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
19467         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
19468         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
19469         adjust buffer in native_interpret_expr call.
19471 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19473         * config/nds32/constants.md (unspec_volatile_element): Add cache
19474         control enum values.
19475         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
19476         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
19477         * config/nds32/nds32.c (nds32_cctl_names): New.
19478         (nds32_print_operand): Handle cache control register names.
19479         * config/nds32/nds32.h (nds32_builtins): New enum values.
19480         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
19481         macros.
19482         * config/nds32/nds32.md (type): Add mmu.
19483         * config/nds32/pipelines.md (simple_insn): Add mmu.
19485 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
19487         * config/nds32/nds32.md (type): Remove call.
19488         * config/nds32/pipelines.md (simple_insn): Likewise.
19490 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19492         * config/nds32/constants.md (unspec_volatile_element): Add
19493         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
19494         UNSPEC_VOLATILE_FMFCFG.
19495         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
19496         description for fmfcfg and fmfcsr.
19497         (bdesc_1arg): Add fmtcsr.
19498         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
19499         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
19500         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
19501         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
19502         unspec_fmfcfg): New patterns.
19503         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
19504         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
19505         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
19506         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
19507         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
19508         __nds32__fmfcfg): Define.
19510 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
19512         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
19513         intrinsic register names.
19514         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
19515         intrinsic register enum values and macros.
19517 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
19519         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
19520         for load/store addressing form.
19521         (nds32_print_operand_address): Likewise.
19523 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
19525         PR target/85196
19526         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
19527         based on LABEL_REF.  Remove useless assertion.
19528         (pic_address_needs_scratch): Fix formatting.
19529         (sparc_legitimize_pic_address): Minor tweaks.
19530         (sparc_delegitimize_address): Adjust assertion accordingly.
19531         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
19532         into symbolic_operand.
19533         (movsi_high_pic_label_ref): Likewise.
19534         (movsi_lo_sum_pic_label_ref): Likewise.
19535         (movdi_pic_label_ref): Likewise.
19536         (movdi_high_pic_label_ref): Likewise.
19537         (movdi_lo_sum_pic_label_ref): Likewise.
19539 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
19541         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
19542         custom LIB_SPEC setup.
19544 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
19545             Kito Cheng  <kito.cheng@gmail.com>
19547         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
19548         * config/riscv/freebsd.h: New.
19550 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
19552         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
19553         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
19554         file.
19556 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
19557             Kito Cheng  <kito.cheng@gmail.com>
19559         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
19560         nds32_output_call, nds32_symbol_binds_local_p): New functions.
19561         * config/nds32/nds32-protos.h (nds32_output_call,
19562         nds32_output_return): Declare.
19563         * config/nds32/nds32.md: Refine all the call and return patterns.
19565 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
19567         PR debug/85252
19568         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
19569         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
19571         PR rtl-optimization/84872
19572         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
19573         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
19574         EDGE_CROSSING edge.
19576 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
19578         * expr.c (copy_blkmode_to_reg): Revert 254862.
19579         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
19581 2018-04-06  Richard Biener  <rguenther@suse.de>
19583         PR middle-end/85244
19584         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
19585         after seeing a component reference with an adjacent field.  Treat
19586         refs to arrays at struct end of external decls similar to
19587         refs to unconstrained commons.
19589 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
19591         PR sanitizer/85213
19592         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
19593         look through SAVE_EXPRs with non-side-effects argument.  Adjust
19594         recursive calls.
19595         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
19596         save_p here.
19598 2018-04-06  Richard Biener  <rguenther@suse.de>
19600         PR middle-end/85180
19601         * alias.c (find_base_term): New wrapper around find_base_term
19602         unwinding CSELIB_VAL_PTR changes.
19603         (find_base_term): Do not restore CSELIB_VAL_PTR during the
19604         recursion.
19606 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19608         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
19609         instructions.
19610         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
19611         constant definitions.
19612         ("nop"): lr 0,0 -> nopr r0
19613         ("nop_lr0", "nop_lr1"): New insn definitions.
19615 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
19617         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
19618         NDS32_V3PUSH_AVAILABLE_P macro.
19620 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
19621             Chung-Ju Wu  <jasonwucj@gmail.com>
19623         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
19624         (nds32*-*-*): Add float and fpu_config into supported_defaults.
19625         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
19626         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
19627         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
19628         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
19629         * config/nds32/constraints.md: New constraints and checking for hard
19630         float configuration.
19631         * config/nds32/iterators.md: New mode iterator and attribute for hard
19632         float configuration.
19633         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
19634         patterns.
19635         * config/nds32/nds32-fpu.md: New file.
19636         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
19637         deal with hard float code generation.
19638         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
19639         ARCH_V3S.
19640         (abi_type, float_reg_number): New enum type.
19641         * config/nds32/nds32-predicates.c: New predicates for hard float.
19642         * config/nds32/nds32-protos.h: Declare functions for hard float.
19643         * config/nds32/nds32.c: Implementation for hard float configuration.
19644         * config/nds32/nds32.h: Definitions for hard float configuration.
19645         * config/nds32/nds32.md: Include hard float machine description and
19646         modify patterns for hard float configuration.
19647         * config/nds32/nds32.opt: New options for hard float configuration.
19648         * config/nds32/predicates.md: New predicates for hard float
19649         configuration.
19651 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
19653         * common/config/nds32/nds32-common.c
19654         (nds32_option_optimization_table): Enable -mreleax-hint by default.
19656 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
19658         PR middle-end/85195
19659         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
19660         CONSTRUCTOR_ELT (ctor, ...)->value.
19662 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
19664         PR target/85193
19665         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
19667 2018-04-05  Tom de Vries  <tom@codesourcery.com>
19669         PR target/85204
19670         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
19671         cond jump.
19673 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
19674             Kito Cheng  <kito.cheng@gmail.com>
19676         * config/nds32/constraints.md (U33): Fine-tune checking condition.
19677         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
19678         * config/nds32/nds32.h (nds32_16bit_address_type): Add
19679         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
19681 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
19682             Kito Cheng  <kito.cheng@gmail.com>
19684         * config/nds32/constraints.md (Ufe): New memory constraint.
19685         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
19686         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
19687         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
19688         operands.
19689         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
19690         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
19692 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19694         * config/nds32/nds32.md: Use optimize_size in the condition for
19695         alu-shift instructions.
19697 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19699         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
19701 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19703         * config/nds32/nds32.md (negsi2): Refine pattern.
19705 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
19706             Chung-Ju Wu  <jasonwucj@gmail.com>
19708         * config/nds32/iterators.md (shift_rotate): New code iterator.
19709         (shift): New code attribute.
19710         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
19711         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
19712         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
19713         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
19714         bit-wise operations.
19715         (andsi3, *andsi3): Ditto.
19716         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
19717         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
19718         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
19719         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
19720         nds32_ior_operand, nds32_xor_operand): New predicates.
19722 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19724         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
19725         (addsi3, subsi3): ... this.
19727 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19729         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
19731 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19733         * config/nds32/nds32.md: Adjust indention.
19735 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
19737         * config/nds32/nds32.md (feature): New attribute.
19739 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
19741         * config/nds32/nds32.md (subtype): New attribute.
19743 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
19745         PR target/85203
19746         * config/arm/arm-builtins.c (arm_expand_builtin): Change
19747         expansion to perform a bitwise AND of the argument followed by a
19748         boolean negation of the result.
19750 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
19752         PR rtl-optimization/84878
19753         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
19754         the basic block.  Assert the use reference is not artificial and that
19755         it has an associated insn.
19757 2018-04-04  Michael Matz  <matz@suse.de>
19759         * builtins.c (compute_objsize): Pass correct operand
19760         to array_at_struct_end_p.
19762 2018-04-04  Richard Biener  <rguenther@suse.de>
19764         PR lto/85176
19765         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
19766         from contexts for DINFO_LEVEL_TERSE and below.
19768 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
19770         * config/nds32/nds32-doubleword.md (move_<mode>): Require
19771         resiter_operand condition.
19772         * config/nds32/nds32.md (*move<mode>): Ditto.
19774 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
19775             Monk Chiang  <sh.chiang04@gmail.com>
19777         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
19779 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19781         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
19783 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19784             Kito Cheng  <kito.cheng@gmail.com>
19786         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
19787         nds32_cond_code_str, output_cond_branch,
19788         output_cond_branch_compare_zero, nds32_expand_cbranch,
19789         nds32_expand_cstore, nds32_expand_movcc,
19790         nds32_output_cbranchsi4_equality_zero,
19791         nds32_output_cbranchsi4_equality_reg,
19792         nds32_output_cbranchsi4_equality_reg_or_const_int,
19793         nds32_output_cbranchsi4_greater_less_zero: New functions.
19794         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
19795         nds32_expand_cstore, nds32_expand_movcc,
19796         nds32_output_cbranchsi4_equality_zero,
19797         nds32_output_cbranchsi4_equality_reg,
19798         nds32_output_cbranchsi4_equality_reg_or_const_int,
19799         nds32_output_cbranchsi4_greater_less_zero): Declare.
19800         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
19801         nds32_rimm11s_operand): New predicates.
19802         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
19803         * config/nds32/nds32.md: Rewrite all the branch and conditional move
19804         patterns.
19806 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
19808         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
19809         * config/nds32/nds32.md: Ditto.
19810         * config/nds32/pipelines.md: Ditto.
19812 2018-04-04  Richard Biener  <rguenther@suse.de>
19814         PR tree-optimization/85168
19815         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
19816         propagating abnormals.
19818 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19820         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
19822 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
19823             Kito Cheng  <kito.cheng@gmail.com>
19825         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
19826         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
19827         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
19828         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
19829         * config/nds32/nds32.md (sibcall_internal): New.
19830         (sibcall_register): Remove.
19831         (sibcall_immediate): Remove.
19832         (sibcall_value_internal): New.
19833         (sibcall_value_register): Remove.
19834         (sibcall_value_immediate): Remove.
19835         * config/nds32/predicates.md (nds32_general_register_operand): New.
19836         (nds32_call_address_operand): New.
19838 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
19840         PR rtl-optimization/85167
19841         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
19842         bb_defs if *split_p, instead preinitialize it to NULL.
19844         PR tree-optimization/85156
19845         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
19846         evaluating the argument multiple times.
19848 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
19850         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
19851         than vector.
19852         (_mm_cvtpd_ps): Likewise.
19853         (_mm_cvttpd_epi32): Likewise.
19854         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
19855         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
19856         vector, pixel, and bool following altivec.h include.
19858 2018-04-03  Martin Sebor  <msebor@redhat.com>
19860         * doc/extend.texi (Common Function Attributes): Clarify.
19861         (const attribute): Likewise.
19862         (pure attribute): Likewise.
19864 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
19866         PR target/85169
19867         * config/i386/i386.c (ix86_expand_vector_set): Use
19868         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
19870 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
19872         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
19873         instructions when changing rounding bits to preserve precision bits
19874         in the x87 control word.
19876 2018-04-03  Martin Liska  <mliska@suse.cz>
19878         PR tree-optimization/82491
19879         * rtl.h (strip_offset_and_add): Replace += suboffset with
19880         poly_uint64 () + suboffset.
19882 2018-03-29  Martin Liska  <mliska@suse.cz>
19883             Martin Jambor  <mjambor@suse.cz>
19885         PR ipa/84947
19886         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
19887         param_type is not an integral or pointer type.
19889 2018-04-03  Richard Biener  <rguenther@suse.de>
19891         * sese.h (recompute_all_dominators): Remove.
19893 2018-04-02  Martin Sebor  <msebor@redhat.com>
19895         * doc/invoke.texi (-Wrestrict): Fix typos.
19897 2018-04-02  Jim Wilson  <jimw@sifive.com>
19899         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
19900         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
19901         (<optab>di3, <optab>si3_extend): Likewise.
19902         (<optab>si3_mask, <optab>si3_mask_1): New.
19903         (<optab>di3_mask, <optab>di3_mask_1): New.
19904         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
19905         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
19906         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
19908 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
19910         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
19911         example.
19913 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
19915         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
19916         (nds32_canonicalize_comparison): New function.
19918 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
19919             Kito Cheng  <kito.cheng@gmail.com>
19920             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
19922         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
19923         * config/nds32/constants.md (unspec_volatile_element): Add
19924         UNSPEC_VOLATILE_RELAX_GROUP.
19925         * config/nds32/nds32-relax-opt.c: New file.
19926         * config/nds32/nds32-predicates.c
19927         (nds32_symbol_load_store_p): New function.
19928         * config/nds32/nds32-protos.h
19929         (nds32_symbol_load_store_p): Declare function.
19930         (make_pass_nds32_relax_opt): Declare new rtl pass function.
19931         * config/nds32/nds32.c
19932         (nds32_register_pass): New function to register pass.
19933         (nds32_register_passes): New function to register passes.
19934         * config/nds32/nds32.md (relax_group): New pattern.
19935         * config/nds32/nds32.opt (mrelax-hint): New option.
19936         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
19938 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
19940         * config/nds32/t-nds32: Modify files dependency.
19942 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
19944         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
19945         (PROFILE_HOOK): Define its implementation.
19947 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
19949         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
19950         type and 32-bit size.
19952 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
19954         PR middle-end/85090
19955         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
19956         (V_128_256): New mode iterator.
19957         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
19958         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
19959         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
19960         of V.
19961         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
19962         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
19964 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
19966         PR target/83315
19967         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
19968         NaN inputs correctly.
19970 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
19972         PR target/80546
19973         * config/rs6000/vsx.md (??r): New mode attribute.
19974         (*vsx_mov<mode>_64bit): Use it.
19975         (*vsx_mov<mode>_32bit): Likewise.
19977 2018-03-30  Martin Sebor  <msebor@redhat.com>
19979         PR tree-optimization/84818
19980         * builtins.c (check_access): Use warning_n.
19982 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
19984         PR target/83822
19985         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
19986         condition.
19987         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
19988         condition.
19990 2018-03-30  Julia Koval  <julia.koval@intel.com>
19992         PR target/84413
19993         * x86-tune.def (movx, partial_reg_dependency): Enable for
19994         m_SKYLAKE_AVX512.
19996 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
19998         PR inline-asm/84985
19999         * lra-constraints.c (process_alt_operands): Move setting
20000         this_alternative_matches below.
20002 2018-03-29  Martin Liska  <mliska@suse.cz>
20004         PR lto/84995.
20005         * doc/invoke.texi: Document how LTO works with debug info.
20006         Describe auto-load support of binutils.  Mention 'x86-64'
20007         as valid option value of -march option.
20009 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
20011         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
20013         PR c/85094
20014         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
20015         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
20016         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
20017         checking.
20019 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
20021         PR target/84912
20022         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
20023         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
20024         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
20025         for RS6000_BTM_POWERPC64.
20026         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
20027         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
20028         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
20029         definition.
20030         (DIVDE): Use it.
20031         (DIVDEU): Likewise.
20033 2018-03-28  Carl Love  <cel@us.ibm.com>
20035         Revert
20036         2017-09-27  Carl Love  <cel@us.ibm.com>
20038         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
20039         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
20040         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
20041         fctiw instruction.
20043 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20045         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
20046         instead of __vector bool.
20047         (_mm_max_pu8): Likewise.
20048         (_mm_min_pi16): Likewise.
20050 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
20052         PR target/84912
20053         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
20054         (DIVWEUO): Likewise.
20055         (DIVDEO): Likewise.
20056         (DIVDEUO): Likewise.
20057         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
20058         DIVWEUO and DIVDEUO.
20059         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
20060         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
20061         (div_extend): Likewise.
20062         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
20063         builtin function.
20064         (__builtin_divweuo): Likewise.
20065         (__builtin_divdeo): Likewise.
20066         (__builtin_divdeuo): Likewise.
20068 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
20070         PR target/85095
20071         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
20072         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
20074         PR tree-optimization/82004
20075         * gimple-match-head.c (optimize_pow_to_exp): New function.
20076         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
20077         Don't fold to exp if optimize_pow_to_exp is false.
20079 2018-03-28  Martin Liska  <mliska@suse.cz>
20081         PR other/84819
20082         * calls.c (initialize_argument_information): Fix trailing space.
20083         * common.opt: Fix typo and provide better explanation for
20084         -fsanitize-coverage option.
20085         * config/i386/i386.opt: Fix typo.
20087 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
20088             Martin Liska  <mliska@suse.cz>
20090         PR sanitizer/85081
20091         * gimplify.c (asan_poison_variable): Don't do the check for
20092         gimplify_omp_ctxp here.
20093         (gimplify_decl_expr): Do it here.
20094         (gimplify_target_expr): Likewise.
20096 2018-03-28  Martin Liska  <mliska@suse.cz>
20098         PR target/84988
20099         * config/i386/i386.c (ix86_function_arg_advance): Do not call
20100         chkp_type_bounds_count if MPX is not enabled.
20102 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
20104         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
20106 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
20108         PR target/84914
20109         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
20110         function to create the function decl for complex long double
20111         multiply and divide for -mabi=ieeelongdouble.
20112         (init_float128_ieee): Call it.
20114 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
20116         PR target/85044
20117         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
20118         -fcf-protection=branch -mibt.
20119         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
20121 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20123         PR target/81863
20124         * config/arm/arm.c (arm_valid_symbolic_address): Handle
20125         arm_word_relocations.
20127 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
20129         PR target/85056
20130         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
20131         extern array declarations.
20133 2018-03-27  Richard Biener  <rguenther@suse.de>
20135         PR middle-end/84067
20136         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
20137         explicit single_use checks.
20139 2018-03-27  Richard Biener  <rguenther@suse.de>
20141         PR tree-optimization/85082
20142         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
20143         Valueize the VUSE.
20145 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
20147         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
20148         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
20149         Turn on fasynchronous-unwind-tables and funwind-tables.
20151 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
20153         PR target/85073
20154         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
20155         (*bmi_blsr_<mode>_ccz): Ditto.
20157 2018-03-26  Tom de Vries  <tom@codesourcery.com>
20159         PR tree-optimization/85063
20160         * omp-general.c (offloading_function_p): New function.  Factor out
20161         of ...
20162         * omp-offload.c (pass_omp_target_link::gate): ... here.
20163         * omp-general.h (offloading_function_p): Declare.
20164         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
20165         with attribute omp declare target for offloading functions.
20167 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
20169         PR tree-optimization/84005
20170         * tree-data-ref.h (get_base_for_alignment): Declare.
20171         * tree-data-ref.c (get_base_for_alignment_1): New function.
20172         (get_base_for_alignment): Likewise.
20173         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
20174         get_base_for_alignment to find a suitable base object, instead
20175         of always using drb->base_address.
20177 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
20179         PR inline-asm/85022
20180         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
20181         known size by default.
20183 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
20185         PR inline-asm/85030
20186         * lra-constraints.c (process_alt_operands): Don't match BLKmode
20187         and non BLKmode operands.
20189 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20191         PR target/85026
20192         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
20193         Clean up attributes.
20195 2018-03-23  Richard Biener  <rguenther@suse.de>
20197         PR debug/85020
20198         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
20199         we are going to emit early debug for LTO.
20201 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
20203         PR inline-asm/85034
20204         * function.c (match_asm_constraints_1): Don't optimize if input
20205         doesn't satisfy general_operand predicate for output's mode.
20207         PR inline-asm/85022
20208         * alias.c (write_dependence_p): Don't require for x_canonicalized
20209         non-VOIDmode if x has VOIDmode.
20211         PR sanitizer/85029
20212         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
20213         just don't try to optimize it rather than assert it never happens.
20215 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20217         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
20218         macro expansions for definition of ST_INTERNAL_<mode> and
20219         LD_INTERNAL_<mode> builtins.
20220         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
20221         Remove prototype.
20222         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
20223         function.
20224         (altivec_expand_st_builtin): Likewise.
20225         (altivec_expand_builtin): Remove calls to deleted functions.
20226         (rs6000_address_for_altivec): Delete this function.
20227         * config/rs6000/vector.md: Remove expands for
20228         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
20230 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
20232         PR target/84826
20233         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
20234         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
20235         re-computing once computed.
20236         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
20237         (arm_init_machine_status): Initialize
20238         machine->static_chain_stack_bytes.
20240 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20242         PR target/84760
20243         * doc/extend.texi: Add four new prototypes for vec_ld.
20244         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
20245         definitions for more logical presentation.
20246         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
20247         entries for V1TI variants of __builtin_altivec_ld builtin.
20248         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
20249         handling of V1TI variant of LVX icode pattern.
20250         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
20251         (rs6000_gimple_fold_builtin): Likewise.
20252         (altivec_init_builtins): Add code to define
20253         __builtin_altivec_lvx_v1ti function.
20255 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
20257         PR inline-asm/84941
20258         * function.c (match_asm_constraints_1): Don't do the optimization
20259         if input isn't a REG, SUBREG, MEM or constant.
20261 2018-03-22  Tom de Vries  <tom@codesourcery.com>
20263         PR tree-optimization/84956
20264         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
20265         bb_has_abnormal_pred.
20267 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
20269         PR sanitizer/85018
20270         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
20271         DECL_INITIAL (decl) to decl at the end.
20272         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
20273         adjust the comment.
20275 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
20277         * doc/extend.texi (__builtin_tgmath): Document when complex
20278         integer types are treated as _Complex _Float64.
20280 2018-03-21  Tom de Vries  <tom@codesourcery.com>
20282         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
20284 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
20286         PR tree-optimization/84960
20287         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
20288         if it is ENTRY block, move them into single succ of ENTRY in that case.
20290 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
20292         PR tree-optimization/84811
20293         * poly-int.h (poly_span_traits): Remove the T3 parameter and
20294         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
20295         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
20296         (known_subrange_p): Update accordingly.  Cast each value involved
20297         in the size comparison, rather than casting the result of the
20298         subtraction.
20300 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
20302         PR tree-optimization/84982
20303         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
20304         by flipping the least significant bit rather than all bits from
20305         bitpos to bitpos + bitsize - 1.
20307 2018-03-21  Nathan Sidwell  <nathan@acm.org>
20309         * doc/extend.texi (Deprecated Features): Remove mention of
20310         long-deleted deprecations.
20312 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20314         PR jit/84288
20315         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
20316         * configure: Regenerate.
20318 2018-03-21  Tom de Vries  <tom@codesourcery.com>
20320         PR tree-optimization/83126
20321         * tree-parloops.c (num_phis): New function.
20322         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
20324 2018-03-21  Nathan Sidwell  <nathan@acm.org>
20326         * doc/extend.texi (Deprecated Features): Update deprecated flags,
20327         mention anon-struct/union members and trailing attributes.
20329 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
20331         PR tree-optimization/84969
20332         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
20333         builtin memset partitions if they set different rhs values.
20335 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
20337         PR rtl-optimization/84989
20338         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
20339         VEC_DUPLICATE with scalar result mode.
20341 2018-03-21  Martin Liska  <mliska@suse.cz>
20343         PR ipa/84963
20344         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
20345         not intended return statement.
20347 2018-03-21  Martin Liska  <mliska@suse.cz>
20349         PR target/84988
20350         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
20351         (chkp_find_bound_slots_1): Limit number of iterations.
20353 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
20355         PR target/84838
20356         * Minor grammar fixes for x86 options.
20358 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
20360         PR debug/84875
20361         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
20362         holding REG_CFA_RESTORE notes, instead turn them into a USE.
20364 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
20366         PR target/83789
20367         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
20368         (altivec_lvx_<mode>_1op): Likewise.
20369         (altivec_stvx_<mode>_2op): Likewise.
20370         (altivec_stvx_<mode>_1op): Likewise.
20371         (altivec_lvx_<VM2:mode>): New define_expand.
20372         (altivec_stvx_<VM2:mode>): Likewise.
20373         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
20374         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
20375         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
20376         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
20377         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
20378         (rs6000_gen_lvx): Likewise.
20379         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
20380         (altivec_expand_stv_builtin): Likewise.
20381         (altivec_expand_builtin): Likewise.
20382         * config/rs6000/vector.md: Likewise.
20384 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20386         PR target/82518
20387         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
20388         BYTES_BIG_ENDIAN.
20390 2018-03-20  Richard Biener  <rguenther@suse.de>
20392         PR target/84986
20393         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
20394         sign-conversions as zero, fall back to standard scalar_stmt
20395         cost for the rest.
20397 2018-03-20  Martin Liska  <mliska@suse.cz>
20399         PR ipa/84825
20400         * predict.c (rebuild_frequencies): Handle case when we have
20401         PROFILE_ABSENT, but flag_guess_branch_prob is false.
20403 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
20405         PR target/84990
20406         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
20407         flag_section_anchors.
20408         * varasm.c (use_blocks_for_decl_p): Remove hack for
20409         dw2_force_const_mem.
20411         PR target/84845
20412         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
20413         to ...
20414         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
20415         be created, use lowpart_subreg of operands[0] rather than operands[0]
20416         itself.
20417         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
20418         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
20419         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
20420         and n constraint instead of aarch64_shift_imm_di and Usd.
20421         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
20422         (*aarch64_<optab>_reg_minus<mode>3): ... this.
20424 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
20426         PR target/82989
20427         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
20428         to favor GPR over NEON registers.
20429         (<shift>di3_neon): Likewise.
20431 2018-03-20  Tom de Vries  <tom@codesourcery.com>
20433         PR target/84952
20434         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
20435         (nvptx_process_pars): Emit bar.sync asap and alap.
20437 2018-03-20  Tom de Vries  <tom@codesourcery.com>
20439         PR target/84954
20440         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
20441         seen_label if seen_label is already set.
20443 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
20445         PR target/84945
20446         * config/i386/i386.c (fold_builtin_cpu): For features above 31
20447         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
20448         Use 1U instead of 1.  Formatting fixes.
20450         PR c/84953
20451         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
20452         instead of TREE_TYPE (s1) for the return value.
20454 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
20456         PR tree-optimization/84946
20457         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
20458         bitsize + bitsize in poly_uint64 rather than poly_int64.
20460         PR sanitizer/78651
20461         * dwarf2asm.c: Include fold-const.c.
20462         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
20463         of decl rather than decl itself.
20465         PR rtl-optimization/84643
20466         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
20468 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
20470         PR sanitizer/78651
20471         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
20472         calling assemble_variable.
20474 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
20476         PR target/81647
20477         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
20478         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
20480 2018-03-19  Jim Wilson  <jimw@sifive.com>
20482         PR bootstrap/84856
20483         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
20484         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
20485         (riscv_first_stack_step): Likewise.
20486         (riscv_option_override): Use STACK_BOUNDARY instead of
20487         MIN_STACK_BOUNDARY.
20488         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
20489         MIN_STACK_BOUNDARY.
20490         (BIGGEST_ALIGNMENT): Set to 128.
20491         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
20492         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
20493         STACK_BOUNDARY.
20495 2018-03-19  Richard Biener  <rguenther@suse.de>
20497         PR tree-optimization/84933
20498         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
20499         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
20501 2018-03-19  Richard Biener  <rguenther@suse.de>
20503         PR tree-optimization/84859
20504         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
20505         (cond_if_else_store_replacement): Perform sinking operation on
20506         single-store BBs regardless of MAX_STORES_TO_SINK setting.
20507         Generalize what a BB with a single eligible store is.
20509 2018-03-19  Richard Biener  <rguenther@suse.de>
20511         PR tree-optimization/84929
20512         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
20513         chrec_is_positive against non-chrec arg.
20515 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
20517         PR target/84711
20518         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
20520 2018-03-18  Martin Liska  <mliska@suse.cz>
20522         PR rtl-optimization/84635
20523         * regrename.c (build_def_use): Use matches_mode only when
20524         matches >= 0.
20526 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
20528         PR tree-optimization/84913
20529         * tree-vect-loop.c (vectorizable_reduction): Don't try to
20530         vectorize chains of COND_EXPRs.
20532 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
20534         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
20536 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
20538         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
20540 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
20542         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
20544 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
20545             Kito Cheng  <kito.cheng@gmail.com>
20547         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
20548         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
20549         (nds32_adjust_reg_alloc_order): New function.
20550         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
20552 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
20554         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
20555         nds32_print_operand, nds32_print_operand_address): Use
20556         HOST_WIDE_INT_PRINT_DEC instead.
20558 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
20560         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
20562 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
20564         PR target/84902
20565         * config/i386/i386.c (initial_ix86_tune_features,
20566         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
20567         unsigned long long.
20568         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
20569         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
20570         rather than 1u << ix86_tune.  Formatting fix.
20571         (ix86_option_override_internal): Change ix86_arch_mask from
20572         unsigned int to unsigned HOST_WIDE_INT, initialize to
20573         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
20574         (ix86_function_specific_restore): Likewise.
20576 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
20578         PR target/84899
20579         * postreload.c (reload_combine_recognize_pattern): Perform
20580         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
20581         truncate_int_for_mode the result for the destination's mode.
20583         PR c/84909
20584         * hsa-gen.c (mem_type_for_type): Fix comment typo.
20585         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
20586         Likewise.
20587         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
20588         Likewise.
20590 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
20592         PR target/84876
20593         * lra-assigns.c (lra_split_hard_reg_for): Don't use
20594         regno_allocno_class_array and sorted_pseudos.
20595         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
20596         insns where regno is used.
20598 2018-03-16  Martin Liska  <mliska@suse.cz>
20600         PR ipa/84833
20601         * multiple_target.c (create_dispatcher_calls): Redirect
20602         reference in the symbol table.
20604 2018-03-16  Martin Liska  <mliska@suse.cz>
20606         PR ipa/84722
20607         * multiple_target.c (create_dispatcher_calls): Redirect also
20608         an alias.
20610 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
20612         PR c++/79937
20613         PR c++/82410
20614         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
20615         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
20616         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
20618 2018-03-16  Julia Koval  <julia.koval@intel.com>
20620         * doc/invoke.texi (Skylake Server): Add CLWB.
20621         Cannonlake): Remove CLWB.
20623 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
20625         PR tree-optimization/84841
20626         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
20627         1 << 3.
20628         (FLOAT_ONE_CONST_TYPE): Define.
20629         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
20630         (sort_by_operand_rank): Put entries with higher constant_type last
20631         rather than first to match comments.
20633 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
20635         * config/nios2/nios2.md (movsi_internal): Fix thinko in
20636         split predicate.
20638 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
20640         PR c++/79085
20641         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
20642         check and use address of target always.
20644 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
20646         PR target/84574
20647         * config/i386/i386.c (indirect_thunk_needed): Update comments.
20648         (indirect_thunk_bnd_needed): Likewise.
20649         (indirect_thunks_used): Likewise.
20650         (indirect_thunks_bnd_used): Likewise.
20651         (indirect_return_needed): New.
20652         (indirect_return_bnd_needed): Likewise.
20653         (output_indirect_thunk_function): Add a bool argument for
20654         function return.
20655         (output_indirect_thunk_function): Don't generate alias for
20656         function return thunk.
20657         (ix86_code_end): Call output_indirect_thunk_function to generate
20658         function return thunks.
20659         (ix86_output_function_return): Set indirect_return_bnd_needed
20660         and indirect_return_needed instead of indirect_thunk_bnd_needed
20661         and indirect_thunk_needed.
20663 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
20665         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
20666         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
20667         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
20669 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
20670             Paul Hua <paul.hua.gm@gmail.com>
20672         PR c/84852
20673         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
20675 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
20677         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
20678         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
20679         resp. SFmode cases.
20681 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
20683         PR target/84711
20684         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
20685         instead of GET_MODE_SIZE when comparing Units.
20687 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
20689         PR target/68256
20690         * varasm.c (hash_section): Return an unchangeble hash value
20691         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
20692         Return !aarch64_can_use_per_function_literal_pools_p ().
20694 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
20696         PR target/84860
20697         * optabs.c (emit_conditional_move): Pass address of cmode's copy
20698         rather than address of cmode as last argument to prepare_cmp_insn.
20700 2018-03-15  Julia Koval  <julia.koval@intel.com>
20702         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
20703         F_AVX512VNNI, F_AVX512BITALG): New.
20705 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
20707         PR target/83451
20708         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
20709         insn for floating-point loads and stores.
20711 2018-03-14  Carl Love  <cel@us.ibm.com>
20713         * config/rs6000/rs6000-c.c: Add macro definitions for
20714         ALTIVEC_BUILTIN_VEC_PERMXOR.
20715         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
20716         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
20717         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
20718         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
20719         UNSPEC_VPERMXOR.
20720         * config/doc/extend.texi: Add prototypes for vec_permxor.
20722 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
20724         PR c/84852
20725         * diagnostic-show-locus.c (class layout_point): Convert m_line
20726         from int to linenum_type.
20727         (line_span::comparator): Use linenum "compare" function when
20728         comparing line numbers.
20729         (test_line_span): New function.
20730         (layout_range::contains_point): Convert param "row" from int to
20731         linenum_type.
20732         (layout_range::intersects_line_p): Likewise.
20733         (layout::will_show_line_p): Likewise.
20734         (layout::print_source_line): Likewise.
20735         (layout::should_print_annotation_line_p): Likewise.
20736         (layout::print_annotation_line): Likewise.
20737         (layout::print_leading_fixits): Likewise.
20738         (layout::annotation_line_showed_range_p): Likewise.
20739         (struct line_corrections): Likewise for field m_row.
20740         (line_corrections::line_corrections): Likewise for param "row".
20741         (layout::print_trailing_fixits): Likewise.
20742         (layout::get_state_at_point): Likewise.
20743         (layout::get_x_bound_for_row): Likewise.
20744         (layout::print_line): Likewise.
20745         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
20746         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
20747         * input.c (selftest::test_linenum_comparisons): New function.
20748         (selftest::input_c_tests): Call it.
20749         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
20750         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
20751         * selftest.h (ASSERT_GT): New macro.
20752         (ASSERT_GT_AT): New macro.
20753         (ASSERT_LT): New macro.
20754         (ASSERT_LT_AT): New macro.
20756 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
20758         PR rtl-optimization/84780
20759         * combine.c (distribute_links): Don't make a link based on pc_rtx.
20761 2018-03-14  Martin Liska  <mliska@suse.cz>
20763         * tree.c (record_node_allocation_statistics): Use
20764         get_stats_node_kind.
20765         (get_stats_node_kind): New function extracted from
20766         record_node_allocation_statistics.
20767         (free_node): Use get_stats_node_kind.
20769 2018-03-14  Richard Biener  <rguenther@suse.de>
20771         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
20772         that the value-set of ANTIC_IN doesn't grow.
20774         Revert
20775         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
20776         member.
20777         (BB_VISITED_WITH_VISITED_SUCCS): New define.
20778         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
20780 2018-03-14  Julia Koval  <julia.koval@intel.com>
20782         * config.gcc (icelake-client, icelake-server): New.
20783         (icelake): Remove.
20784         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
20785         (initial_ix86_arch_features): Ditto.
20786         (PTA_SKYLAKE): Add SGX.
20787         (PTA_ICELAKE): Remove.
20788         (PTA_ICELAKE_CLIENT): New.
20789         (PTA_ICELAKE_SERVER): New.
20790         (ix86_option_override_internal): Split up icelake on icelake client and
20791         icelake server.
20792         (get_builtin_code_for_version): Ditto.
20793         (fold_builtin_cpu): Ditto.
20794         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
20795         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
20796         * config/i386/i386.h (processor_type): Ditto.
20797         * doc/invoke.texi: Ditto.
20799 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
20801         PR sanitizer/83392
20802         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
20803         INTEGER_CST offset, add it together with bitpos / 8 and
20804         sign extend based on POINTER_SIZE.
20806         PR target/84844
20807         Revert
20808         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
20810         PR target/78090
20811         * config/i386/constraints.md (Yc): New register constraint.
20812         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
20813         Use Yc constraint for alternative 2 of operand 0.  Remove
20814         preferred_for_speed attribute.
20816 2018-03-14  Richard Biener  <rguenther@suse.de>
20818         PR tree-optimization/84830
20819         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
20820         with the old one to avoid oscillations.
20822 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
20824         PR target/83712
20825         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
20826         pseudos.
20827         (assign_by_spills): Return a flag of reload assignment failure.
20828         Do not process the reload assignment failures.  Do not spill other
20829         reload pseudos if they has the same reg class.  Update n if
20830         necessary.
20831         (lra_assign): Add a return arg.  Set up from the result of
20832         assign_by_spills call.
20833         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
20834         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
20835         usage_insns if it is not NULL.
20836         (spill_hard_reg_in_range): New function.
20837         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
20838         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
20839         function prototypes.
20840         (lra_assign): Change prototype.
20841         * lra.c (lra): Add code to deal with fails by splitting hard reg
20842         live ranges.
20844 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
20846         * config/riscv/riscv.opt (mrelax): New option.
20847         * config/riscv/riscv.c (riscv_file_start): Emit ".option
20848         "norelax" when riscv_mrelax is disabled.
20849         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
20851 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
20853         PR target/84743
20854         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
20855         reassociation for int modes.
20857 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
20859         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
20860         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
20861         for big-endian.
20862         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
20863         * config/aarch64/aarch64-sve.md
20864         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
20865         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
20866         (*extend<mode><Vwide>2): Rename to...
20867         (aarch64_sve_extend<mode><Vwide>2): ...this.
20868         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
20869         renaming the old pattern to...
20870         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
20871         unsigned packs.
20872         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
20873         define_expand, renaming the old pattern to...
20874         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
20875         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
20876         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
20877         account when deciding which SVE instruction the optab should use.
20878         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
20880 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
20882         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
20883         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
20884         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
20885         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
20886         (tlsdesc_small_<mode>): Turn a define_expand and use
20887         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
20888         (tlsdesc_small_advsimd_<mode>): ...this.
20889         (tlsdesc_small_sve_<mode>): New pattern.
20891 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
20893         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
20894         (UNSPEC_UMUL_HIGHPART): New constants.
20895         (MUL_HIGHPART): New int iteraor.
20896         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
20897         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
20898         define_expand.
20899         (*<su>mul<mode>3_highpart): New define_insn.
20901 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
20903         PR lto/84805
20904         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
20905         incomplete types.
20907 2018-03-13  Martin Liska  <mliska@suse.cz>
20909         PR ipa/84658.
20910         * (sem_item_optimizer::sem_item_optimizer): Initialize new
20911         vector.
20912         (sem_item_optimizer::~sem_item_optimizer): Release it.
20913         (sem_item_optimizer::merge_classes): Register variable aliases.
20914         (sem_item_optimizer::fixup_pt_set): New function.
20915         (sem_item_optimizer::fixup_points_to_sets): Likewise.
20916         * ipa-icf.h: Declare new variables and functions.
20918 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
20920         PR middle-end/84834
20921         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
20922         integer_pow2p@2 and test integer_pow2p in condition.
20923         (A < 0 ? C : 0): Similarly for @1.
20925         PR middle-end/84831
20926         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
20927         characters starting at p contain '\0' character, don't look beyond
20928         that.
20930         PR target/84827
20931         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
20932         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
20934         PR target/84828
20935         * reg-stack.c (change_stack): Change update_end var from int to
20936         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
20937         also call set_block_for_insn on the newly added insns and rescan.
20939         PR target/84786
20940         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
20941         on the last operand.
20943         PR c++/84704
20944         * tree.c (stabilize_reference_1): Return save_expr (e) for
20945         STATEMENT_LIST even if it doesn't have side-effects.
20947 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
20949         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
20951 2018-03-12  Renlin Li  <renlin.li@arm.com>
20953         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
20954         aarch64_output_scalar_simd_mov_immediate.
20956 2018-03-12  Martin Sebor  <msebor@redhat.com>
20958         PR tree-optimization/83456
20959         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
20960         for perfectly overlapping calls to memcpy.
20961         (gimple_fold_builtin_memory_chk): Same.
20962         (gimple_fold_builtin_strcpy): Handle no-warning.
20963         (gimple_fold_builtin_stxcpy_chk): Same.
20964         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
20966 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
20968         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
20969         parameter.  Use it for SFmode.
20970         (rs6000_function_arg_advance_1): Adjust.
20971         (rs6000_function_arg): Adjust.
20972         (rs6000_gimplify_va_arg): Pass false for that new parameter.
20974 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
20976         PR rtl-optimization/84169
20977         PR rtl-optimization/84780
20978         * combine.c (can_combine_p): Check for a 2-insn combination whether
20979         the destination register is used between the two insns, too.
20981 2018-03-12  Richard Biener  <rguenther@suse.de>
20983         PR tree-optimization/84803
20984         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
20985         for refs DR analysis didn't process.
20987 2018-03-12  Richard Biener  <rguenther@suse.de>
20989         PR tree-optimization/84777
20990         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
20991         force-vectorize loops ignore whether we are optimizing for size.
20993 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
20995         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
20996         (TARGET_MD_ASM_ADJUST): Define.
20998 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
20999             Kito Cheng  <kito.cheng@gmail.com>
21000             Chung-Ju Wu  <jasonwucj@gmail.com>
21002         * config/nds32/nds32.c (nds32_compute_stack_frame,
21003         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
21004         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
21005         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
21006         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
21007         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
21008         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
21009         * config/nds32/nds32.md (prologue, epilogue): Use macro
21010         NDS32_V3PUSH_AVAILABLE_P to do checking.
21012 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
21014         PR debug/58150
21015         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
21016         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
21017         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
21018         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
21019         addition of most attributes on !orig_type_die or the attribute not
21020         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
21022 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21023             Chung-Ju Wu  <jasonwucj@gmail.com>
21025         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
21026         __NDS32_VH__ macro.
21027         * config/nds32/nds32.opt (mvh): New option.
21029 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21030             Chung-Ju Wu  <jasonwucj@gmail.com>
21032         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
21033         function.
21034         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
21035         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
21036         definition.
21038 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
21039             Chung-Ju Wu  <jasonwucj@gmail.com>
21041         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
21042         function.
21043         * config/nds32/nds32-multiple.md (strlensi): New pattern.
21044         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
21046 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
21047             Kito Cheng  <kito.cheng@gmail.com>
21048             Chung-Ju Wu  <jasonwucj@gmail.com>
21050         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
21051         UNSPEC_FFMISM and UNSPEC_FLMISM.
21052         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
21053         for ffb, ffmism and flmism.
21054         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
21055         (unspec_ffmism): Ditto.
21056         (unspec_flmism): Ditto.
21057         (nds32_expand_builtin_impl): Check if string extension is available.
21058         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
21059         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
21061 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
21063         Reverting patch:
21064         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
21066         PR target/83712
21067         * lra-assigns.c (assign_by_spills): Return a flag of reload
21068         assignment failure.  Do not process the reload assignment
21069         failures.  Do not spill other reload pseudos if they has the same
21070         reg class.
21071         (lra_assign): Add a return arg.  Set up from the result of
21072         assign_by_spills call.
21073         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21074         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21075         usage_insns if it is not NULL.
21076         (spill_hard_reg_in_range): New function.
21077         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21078         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21079         function prototypes.
21080         (lra_assign): Change prototype.
21081         * lra.c (lra): Add code to deal with fails by splitting hard reg
21082         live ranges.
21084 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
21086         PR target/84807
21087         * config/i386/i386.opt: Replace Enforcment with Enforcement.
21089 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
21091         PR debug/84620
21092         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
21093         (dw_val_node): Add val_symbolic_view.
21094         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
21095         (symview_upper_bound): New.
21096         (new_line_info_table): Initialize symviews_since_reset.
21097         (dwarf2out_source_line): Count symviews_since_reset and set
21098         symview_upper_bound.
21099         (dw_val_equal_p): Handle symview.
21100         (add_AT_symview): New.
21101         (print_dw_val): Handle symview.
21102         (attr_checksum, attr_checksum_ordered): Likewise.
21103         (same_dw_val_p, size_of_die): Likewise.
21104         (value_format, output_die): Likewise.
21105         (add_high_low_attributes): Use add_AT_symview for entry_view.
21106         (dwarf2out_finish): Reset symview_upper_bound, clear
21107         zero_view_p.
21109 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
21111         PR target/83969
21112         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
21113         Add strict argument and use it.
21114         (rs6000_split_multireg_move): Update for new strict argument.
21115         (mem_operand_gpr): Disallow all non-offsettable addresses.
21116         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
21118 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
21120         PR target/84772
21121         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
21122         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
21123         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
21125         PR c++/84767
21126         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
21127         decl, use remap_type if we want to use the type.
21129 2018-03-09  Martin Sebor  <msebor@redhat.com>
21131         PR tree-optimization/84526
21132         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
21133         Remove dead code.
21134         (builtin_access::generic_overlap): Be prepared to handle non-array
21135         base objects.
21137 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
21139         PR rtl-optimization/84682
21140         * lra-constraints.c (process_address_1): Check is_address flag
21141         for address constraints.
21142         (process_alt_operands): Likewise.
21143         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
21144         preprocess_constraints.
21145         * recog.h (preprocess_constraints): Add oploc parameter.
21146         Adjust callers.
21147         * recog.c (preprocess_constraints): Test address_operand for
21148         CT_ADDRESS constraints.
21150 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
21152         PR target/83712
21153         * lra-assigns.c (assign_by_spills): Return a flag of reload
21154         assignment failure.  Do not process the reload assignment
21155         failures.  Do not spill other reload pseudos if they has the same
21156         reg class.
21157         (lra_assign): Add a return arg.  Set up from the result of
21158         assign_by_spills call.
21159         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
21160         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
21161         usage_insns if it is not NULL.
21162         (spill_hard_reg_in_range): New function.
21163         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
21164         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
21165         function prototypes.
21166         (lra_assign): Change prototype.
21167         * lra.c (lra): Add code to deal with fails by splitting hard reg
21168         live ranges.
21170 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21172         PR target/83193
21173         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
21174         Accept complain bool parameter.  Only emit errors if it is true.
21175         (arm_parse_cpu_option_name): Likewise.
21176         (arm_target_thumb_only): Adjust callers of the above.
21177         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
21178         prototype to take a default true bool parameter.
21179         (arm_parse_arch_option_name): Likewise.
21181 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
21182             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
21184         PR jit/64089
21185         PR jit/84288
21186         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
21187         * configure: Regenerate.
21188         * configure.ac ("linker --version-script option"): New.
21189         ("linker soname option"): New.
21191 2018-03-09  Richard Biener  <rguenther@suse.de>
21193         PR tree-optimization/84775
21194         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
21195         immediate uses of predicate stmts and mark them modified.
21197         Revert
21198         PR tree-optimization/84178
21199         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
21200         to caller.
21201         (version_loop_for_if_conversion): Delay update_ssa call.
21202         (tree_if_conversion): Delay update_ssa until after predicate
21203         insertion.
21205 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
21207         PR target/84763
21208         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
21209         when the function accesses prior frames.
21211 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
21213         PR debug/84456
21214         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
21215         gen_llsym, otherwise call maybe_gen_llsym.
21217         PR inline-asm/84742
21218         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
21219         has ',' character inside of it.
21221 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21223         PR target/84748
21224         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
21225         as clobbering CC_REGNUM.
21227 2018-03-08  Richard Biener  <rguenther@suse.de>
21229         PR middle-end/84552
21230         * tree-scalar-evolution.c: Include tree-into-ssa.h.
21231         (follow_copies_to_constant): Do not follow SSA names registered
21232         for update.
21234 2018-03-08  Richard Biener  <rguenther@suse.de>
21236         PR tree-optimization/84178
21237         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
21238         to caller.
21239         (version_loop_for_if_conversion): Delay update_ssa call.
21240         (tree_if_conversion): Delay update_ssa until after predicate
21241         insertion.
21243 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
21245         PR tree-optimization/84178
21246         * tree-if-conv.c (release_bb_predicate): Remove the
21247         the assertion that the stmts have NULL use_ops.
21248         Discard the statements, asserting that they haven't
21249         yet been added to a BB.
21251 2018-03-08  Richard Biener  <rguenther@suse.de>
21253         PR tree-optimization/84746
21254         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
21255         (phi_translate): Pass in destination ANTIC_OUT set.
21256         (phi_translate_1): Likewise.  For a simplified result lookup
21257         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
21258         (phi_translate_set): Adjust.
21259         (do_pre_regular_insertion): Likewise.
21260         (do_pre_partial_partial_insertion): Likewise.
21262 2018-03-08  Martin Liska  <mliska@suse.cz>
21264         PR gcov-profile/84735
21265         * doc/gcov.texi: Document usage of profile files.
21266         * gcov-io.h: Document changes in the format.
21268 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
21270         PR debug/84404
21271         PR debug/84408
21272         * dwarf2out.c (struct dw_line_info_table): Update comments for
21273         view == -1.
21274         (FORCE_RESET_NEXT_VIEW): New.
21275         (FORCE_RESETTING_VIEW_P): New.
21276         (RESETTING_VIEW_P): Check for -1 too.
21277         (ZERO_VIEW_P): Likewise.
21278         (new_line_info_table): Force-reset next view.
21279         (dwarf2out_begin_function): Likewise.
21280         (dwarf2out_source_line): Simplify zero_view_p initialization.
21281         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
21282         view directly.  Omit view when omitting .loc at line 0.
21284 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
21286         PR tree-optimization/84740
21287         * tree-switch-conversion.c (process_switch): Call build_constructors
21288         only if info.phi_count is non-zero.
21290         PR tree-optimization/84739
21291         * tree-tailcall.c (find_tail_calls): Check call arguments against
21292         DECL_ARGUMENTS (current_function_decl) rather than
21293         DECL_ARGUMENTS (func) when checking for tail recursion.
21295 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
21297         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
21298         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
21299         Volker Reichelt's entry and add entries for people that perform
21300         GCC fuzzy testing and report numerous bugs.
21302 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
21304         PR target/82411
21305         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
21306         readonly data in sdata, if that is disabled.
21307         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
21308         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
21309         -mreadonly-in-sdata option.
21311 2018-03-07  Martin Sebor  <msebor@redhat.com>
21313         PR tree-optimization/84468
21314         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
21315         basic block when looking for nul assignment.
21317 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
21319         PR target/84277
21320         * except.h (output_function_exception_table): Adjust prototype.
21321         * except.c (output_function_exception_table): Remove FNNAME parameter
21322         and add SECTION parameter.  Ouput one part of the table at a time.
21323         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
21324         the first part of the exception table and emit unwind directives.
21325         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
21326         (i386_pe_seh_cold_init): Likewise.
21327         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
21328         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
21329         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
21330         (ix86_output_call_insn): Emit a nop in one more case for SEH.
21331         * config/i386/winnt.c: Include except.h.
21332         (struct seh_frame_state): Add reg_offset, after_prologue and
21333         in_cold_section fields.
21334         (i386_pe_seh_end_prologue): Set seh->after_prologue.
21335         (i386_pe_seh_cold_init): New function.
21336         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
21337         to seh->in_cold_section.
21338         (seh_emit_push): Record the offset of the push.
21339         (seh_emit_save): Record the offet of the save.
21340         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
21341         Test seh->after_prologue to disregard the epilogue.
21342         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
21343         (i386_pe_end_cold_function): New function.
21345 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
21347         PR fortran/84565
21348         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
21349         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
21351         PR c++/84704
21352         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
21353         on tmp_var.
21354         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
21355         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
21357         PR middle-end/84723
21358         * multiple_target.c: Include tree-inline.h and intl.h.
21359         (expand_target_clones): Diagnose and fail if node->definition and
21360         !tree_versionable_function_p (node->decl).
21362 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
21364         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
21365         sprint_ul.
21366         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
21367         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
21368         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
21370 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
21372         PR target/84710
21373         * combine.c (try_combine): Use reg_or_subregno instead of handling
21374         just paradoxical SUBREGs and REGs.
21376 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
21378         * config/arc/arc.c (arc_finalize_pic): Remove function.
21379         (arc_must_save_register): We use single base PIC register, remove
21380         checks to save/restore the PIC register.
21381         (arc_expand_prologue): Likewise.
21382         * config/arc/arc-protos.h (arc_set_default_type_attributes):
21383         Remove.
21384         (arc_verify_short): Likewise.
21385         (arc_attr_type): Likewise.
21386         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
21387         (walk_stores): Likewise.
21388         (arc_address_cost): Make it static.
21389         (arc_verify_short): Likewise.
21390         (branch_dest): Likewise.
21391         (arc_attr_type): Likewise.
21392         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
21393         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
21394         (arc_final_prescan_insn): Remove inserting the nops due to
21395         hardware hazards.  It is done in reorg step.
21396         (insn_length_variant_t): Remove.
21397         (insn_length_parameters_t): Likewise.
21398         (arc_insn_length_parameters): Likewise.
21399         (arc_get_insn_variants): Likewise.
21400         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
21402 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
21404         PR inline-asm/84683
21405         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
21406         assertion failure.
21408         PR tree-optimization/84687
21409         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
21410         on new_node->decl.
21411         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
21413 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21415         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
21416         Rename to ppc_speculation_barrier.
21417         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
21418         __builtin_ppc_speculation_barrier.
21420 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
21422         PR target/84700
21423         * combine.c (combine_simplify_rtx): Don't try to simplify if
21424         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
21425         are equal to x.
21427 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
21429         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
21430         to 32 bytes when compiling for POWER9.
21432 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
21434         PR target/84564
21435         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
21436         regparm >= 3 with no arg reg available also for calls with
21437         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
21439         PR target/84524
21440         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
21441         orig,vex.
21442         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
21444 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
21446         PR target/84264
21447         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
21449 2018-03-05  Richard Biener  <rguenther@suse.de>
21451         PR tree-optimization/84486
21452         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
21453         When inserting a __builtin_assume_aligned call set the LHS
21454         SSA name alignment info accordingly.
21456 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
21458         PR tree-optimization/84114
21459         * config/aarch64/aarch64.c (aarch64_reassociation_width)
21460         Avoid reassociation of FLOAT_MODE addition.
21462 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
21464         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
21465         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
21466         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
21467         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
21468         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
21469         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
21470         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
21471         and -mwbnoinvd.
21472         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
21473         __builtin_ia32_wbinvd): New builtins.
21474         (SPECIAL_ARGS2): New.
21475         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
21476         (SPECIAL_ARGS2): New.
21477         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
21478         (ix86_valid_target_attribute_inner_p): Ditto.
21479         (ix86_init_mmx_sse_builtins): Add special_args2.
21480         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
21481         TARGET_WBNOINVD_P): New.
21482         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
21483         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
21484         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
21485         * config/i386/immintrin.h (_wbinvd): New intrinsic.
21486         * config/i386/pconfigintrin.h: New file.
21487         * config/i386/wbnoinvdintrin.h: Ditto.
21488         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
21489         wbnoinvdintrin.h.
21490         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
21492 2018-03-05  Richard Biener  <rguenther@suse.de>
21494         PR tree-optimization/84670
21495         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
21496         member.
21497         (BB_VISITED_WITH_VISITED_SUCCS): New define.
21498         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
21499         (compute_antic_aux): Only assert the number of values in ANTIC_IN
21500         doesn't grow if all successors (recursively) were visited at least
21501         once.
21503 2018-03-05  Richard Biener  <rguenther@suse.de>
21505         PR tree-optimization/84650
21506         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
21507         if executed in the loop pipeline.
21509 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
21511         * doc/configfiles.texi (Configuration Files): Move info about
21512         conditionalizing $target-protos.h to...
21513         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
21514         differs from $target-protos.h.
21516 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
21517             Chung-Ju Wu  <jasonwucj@gmail.com>
21519         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
21520         * config/nds32/nds32-multiple.md (setmemsi): Define.
21521         * config/nds32/nds32-memory-manipulation.c
21522         (nds32_gen_dup_4_byte_to_word_value): New.
21523         (emit_setmem_word_loop): New.
21524         (emit_setmem_byte_loop): New.
21525         (nds32_expand_setmem_loop): New.
21526         (nds32_expand_setmem_loop_v3m): New.
21527         (nds32_expand_setmem_unroll): New.
21528         (nds32_expand_setmem): New.
21530 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
21531             Chung-Ju Wu  <jasonwucj@gmail.com>
21533         * config/nds32/nds32-memory-manipulation.c
21534         (nds32_emit_load_store): New.
21535         (nds32_emit_post_inc_load_store): New.
21536         (nds32_emit_mem_move): New.
21537         (nds32_emit_mem_move_block): New.
21538         (nds32_expand_movmemsi_loop_unknown_size): New.
21539         (nds32_expand_movmemsi_loop_known_size): New.
21540         (nds32_expand_movmemsi_loop): New.
21541         (nds32_expand_movmemsi_unroll): New.
21542         (nds32_expand_movmemqi): Rename ...
21543         (nds32_expand_movmemsi): ... to this.
21544         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
21545         (movmemsi): ... to this.
21546         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
21547         (nds32_expand_movmemsi): ... to this.
21549 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
21550             Monk Chiang  <sh.chiang04@gmail.com>
21551             Chung-Ju Wu  <jasonwucj@gmail.com>
21553         * config/nds32/nds32-protos.h
21554         (nds32_expand_load_multiple): New arguments.
21555         (nds32_expand_store_multiple): Ditto.
21556         (nds32_valid_multiple_load_store): Rename ...
21557         (nds32_valid_multiple_load_store_p): ... to this.
21558         * config/nds32/nds32-memory-manipulation.c
21559         (nds32_expand_load_multiple): Refine implementation.
21560         (nds32_expand_store_multiple): Ditto.
21561         * config/nds32/nds32-multiple.md
21562         (load_multiple): Update nds32_expand_load_multiple interface.
21563         (store_multiple): Update nds32_expand_store_multiple interface.
21564         * config/nds32/nds32-predicates.c
21565         (nds32_valid_multiple_load_store): Rename ...
21566         (nds32_valid_multiple_load_store_p): ... to this and refine
21567         implementation.
21568         * config/nds32/predicates.md
21569         (nds32_load_multiple_and_update_address_operation): New predicate.
21570         (nds32_store_multiple_and_update_address_operation): New predicate.
21572 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
21573             Chung-Ju Wu  <jasonwucj@gmail.com>
21575         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
21576         (combo): New attribute.
21577         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
21579 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
21581         * config/nds32/nds32.opt: Change -mcmodel= default value.
21583 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
21584             Monk Chiang  <sh.chiang04@gmail.com>
21585             Chung-Ju Wu  <jasonwucj@gmail.com>
21587         * config/nds32/constants.md (unspec_element): New enum.
21588         * config/nds32/constraints.md (Umw): New constraint.
21589         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
21590         * config/nds32/nds32-intrinsic.md: Likewise.
21591         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
21592         (nds32_valid_smw_lwm_base_p): New.
21593         (nds32_output_smw_single_word): New.
21594         (nds32_output_lmw_single_word): New.
21595         (nds32_expand_unaligned_load): New.
21596         (nds32_expand_unaligned_store): New.
21597         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
21598         (nds32_output_smw_single_word): Declare.
21599         (nds32_output_lmw_single_word): Declare.
21600         (nds32_expand_unaligned_load): Declare.
21601         (nds32_expand_unaligned_store): Declare.
21602         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
21603         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
21604         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
21605         NDS32_BUILTIN_UASTORE_DW.
21606         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
21607         predicate.
21609 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
21610             Kito Cheng  <kito.cheng@gmail.com>
21611             Chung-Ju Wu  <jasonwucj@gmail.com>
21613         * config/nds32/nds32-intrinsic.c
21614         (nds32_expand_builtin_null_ftype_reg): Delete.
21615         (nds32_expand_builtin_reg_ftype_imm): Ditto.
21616         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
21617         (nds32_read_argument): New.
21618         (nds32_legitimize_target): Ditto.
21619         (nds32_legitimize_argument): Ditto.
21620         (nds32_check_constant_argument): Ditto.
21621         (nds32_expand_unop_builtin): Ditto.
21622         (nds32_expand_unopimm_builtin): Ditto.
21623         (nds32_expand_binop_builtin): Ditto.
21624         (nds32_builtin_decl_impl): Ditto.
21625         (builtin_description): Ditto.
21626         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
21627         (nds32_init_builtins_impl): Ditto.
21628         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
21629         (nds32_builtin_decl): New.
21630         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
21631         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
21633 2018-03-02  Jeff Law  <law@redhat.com>
21635         * reorg.c (stop_search_p): Handle DEBUG_INSN.
21636         (redundant_insn, fill_simple_delay_slots): Likewise.
21637         (fill_slots_from_thread): Likewise.
21638         * resource.c (mark_referenced_resources): Likewise.
21639         (mark_set_resources, find_dead_or_set_registers): Likewise.
21641 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21643         * substring-locations.h (format_warning_va): Formatting fix for
21644         ATTRIBUTE_GCC_DIAG.
21645         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
21646         argument.
21647         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
21648         * substring-locations.c: Include intl.h.
21649         (format_warning_va): Turned into small wrapper around
21650         format_warning_n_va, renamed to ...
21651         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
21652         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
21653         use ngettext.
21654         (format_warning_at_substring_n): New function.
21655         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
21656         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
21657         format_warning_at_substring with just a shorter name instead of
21658         const function pointer.
21659         (fmtwarn_n): New function.
21660         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
21661         appropriate, get rid of all the fmtstr temporaries, move conditionals
21662         with G_() wrapped string literals directly into fmtwarn arguments,
21663         cast dir.len to (int), formatting fixes.
21665 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
21667         * doc/invoke.texi: Remove "Cilk Plus" references.
21669 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21670             Richard Biener  <rguenther@suse.de>
21672         PR ipa/84628
21673         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
21674         for error or warning attributes if CALL_FROM_THUNK_P is set.
21675         Formatting fixes.
21677 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21679         PR target/56540
21680         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
21681         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
21683         PR target/56540
21684         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
21685         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
21687         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
21688         instead of -1U in last predictors element's probability member.
21690 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
21692         PR ipa/83983
21693         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21694         arguments if they are comparable.
21696 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
21698         PR tree-optimization/84634
21699         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
21700         masks and masked_loop_p with a single loop_masks, making sure it's
21701         null for bb vectorization.
21703 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
21705         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
21706         (vect_analyze_data_ref_access): Use loop->safe_len rather than
21707         loop->force_vectorize to check whether there is no alias.
21709 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
21711         PR target/84614
21712         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
21713         prototypes.
21714         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
21715         comments.
21716         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
21717         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
21718         instead of a loop around prev_real_insn.
21719         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
21720         prev_real_insn.
21722         PR inline-asm/84625
21723         * config/i386/i386.c (ix86_print_operand): Use conditional
21724         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
21725         zero vector.
21727 2018-03-02  Richard Biener  <rguenther@suse.de>
21729         PR tree-optimization/84427
21730         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
21731         (bitmap_set_subtract_values): Rewrite to handle multiple
21732         exprs per value.
21733         (clean): Likewise.
21734         (prune_clobbered_mems): Likewise.
21735         (phi_translate): Take edge instead of pred/phiblock.
21736         (phi_translate_1): Likewise.
21737         (phi_translate_set): Likewise.  Insert all translated
21738         exprs for a value into the set, keeping possibly multiple
21739         expressions per value.
21740         (compute_antic_aux): Adjust for phi_translate changes.
21741         When intersecting union the expressions and prune those
21742         not in the final value set, keeping possibly multiple
21743         expressions per value.  Do not use value-insertion
21744         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
21745         all expressions.  Add verification that the value-sets
21746         only shrink during iteration.
21747         (compute_partial_antic_aux): Adjust for the phi_translate changes.
21748         (do_pre_regular_insertion): Likewise.
21749         (do_pre_partial_partial_insertion): Likewise.
21751 2018-03-02  Richard Biener  <rguenther@suse.de>
21753         PR target/82005
21754         * config/darwin.c (saved_debug_info_level): New static global.
21755         (darwin_asm_lto_start): Disable debug info generation for LTO out.
21756         (darwin_asm_lto_end): Restore debug info generation settings.
21758 2018-03-01  Martin Liska  <mliska@suse.cz>
21760         PR sanitizer/82484
21761         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
21762         volatile arguments.
21764 2018-03-01  Richard Biener  <rguenther@suse.de>
21766         PR debug/84645
21767         * dwarf2out.c (gen_variable_die): Properly handle late VLA
21768         type annotation with LTO when debug was disabled at compile-time.
21770 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
21772         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
21773         XINT with INTVAL.
21774         (mips_final_postscan_insn): Likewise.
21776 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
21778         PR rtl-optimization/84528
21779         * alias.c (init_alias_target): Add commentary.
21780         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
21781         a unique base value if the frame pointer is not eliminated
21782         to the stack pointer.
21784 2018-03-01  Tom de Vries  <tom@codesourcery.com>
21786         PR rtl-optimization/83327
21787         * lra-int.h (hard_regs_spilled_into): Declare.
21788         * lra.c (hard_regs_spilled_into): Define.
21789         (init_reg_info): Init hard_regs_spilled_into.
21790         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
21791         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
21792         (process_bb_lives): Handle hard_regs_spilled_into.
21793         (lra_create_live_ranges_1): Before doing liveness propagation, clear
21794         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
21796 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
21798         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
21799         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
21800         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
21801         * config/rs6000/aix72.h: New file.
21803 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
21805         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
21806         instead of warning_at with conditional singular and plural messages
21807         where possible.
21809         PR target/52991
21810         * stor-layout.c (update_alignment_for_field): For
21811         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
21812         && !DECL_PACKED (field), do the alignment update, just use
21813         only desired_align instead of MAX (type_align, desired_align)
21814         as the alignment.
21815         (place_field): Don't do known_align < desired_align handling
21816         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
21817         is non-NULL, instead do it after rli->prev_field handling and
21818         only if not within a bitfield word.  For DECL_PACKED (field)
21819         use type_align of BITS_PER_UNIT.
21821 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
21823         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
21824         superfluous parentheses and trailing spaces.
21826 2018-02-28  Richard Biener  <rguenther@suse.de>
21828         PR tree-optimization/84584
21829         * graphite-scop-detection.c (scop_detection::add_scop): Discard
21830         SCoPs with fake exit edge.
21832 2018-02-28  Martin Liska  <mliska@suse.cz>
21834         PR testsuite/84597
21835         * timevar.c (timer::print): Fix format to properly print 100%
21836         values.
21838 2018-02-28  Richard Biener  <rguenther@suse.de>
21840         PR middle-end/84607
21841         * genmatch.c (capture_info::walk_match): Do not mark
21842         captured expressions without operands as expr_p given
21843         they act more like predicates and should be subject to
21844         "lost tail" side-effect preserving.
21846 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
21848         PR rtl-optimization/81611
21849         * auto-inc-dec.c (attempt_change): Move dead note from
21850         mem_insn if it's the next use of regno
21851         (find_address): Take address use of reg holding
21852         non-incremented value.  Add parm to limit search to the named
21853         reg only.
21854         (merge_in_block): Attempt to use a mem insn that is the next
21855         use of the original regno.
21857 2018-02-27  Martin Sebor  <msebor@redhat.com>
21859         PR c++/83871
21860         * doc/invoke.texi (-Wmissing-attributes): New option.
21861         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
21863 2018-02-27  Martin Sebor  <msebor@redhat.com>
21865         PR translation/84207
21866         * diagnostic-core.h (warning_n, error_n, inform_n): Change
21867         n argument to unsigned HOST_WIDE_INT.
21868         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
21869         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
21870         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
21871         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
21873 2018-02-27  Richard Biener  <rguenther@suse.de>
21875         PR tree-optimization/84512
21876         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
21877         Do not use the estimate returned from record_stmt_cost for
21878         the scalar iteration cost but sum properly using add_stmt_cost.
21880 2018-02-27  Richard Biener  <rguenther@suse.de>
21882         PR tree-optimization/84466
21883         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
21884         Adjust last change to less strictly validate use operands.
21886 2018-02-27  Martin Liska  <mliska@suse.cz>
21888         PR gcov-profile/84548
21889         * gcov.c (process_file): Allow partial overlap and consider it
21890         also as group functions.
21891         (output_lines): Properly calculate range of lines for a group.
21893 2018-02-27  Martin Liska  <mliska@suse.cz>
21895         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
21896         'ggc' suffixes.  Change first column width.
21897         (timer::print): Fix formatting of the column.
21899 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
21901         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
21902         preserve inline entry blocks for the sake of debug inline
21903         entry point markers alone.
21904         (remove_unused_locals): Suggest in comments a better place to
21905         force the preservation of inline entry blocks that are
21906         otherwise unused, but do not preserve them.
21908 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
21910         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
21912 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
21914         PR target/84039
21915         * config/i386/constraints.md (Bs): Replace
21916         ix86_indirect_branch_register with
21917         TARGET_INDIRECT_BRANCH_REGISTER.
21918         (Bw): Likewise.
21919         * config/i386/i386.md (indirect_jump): Likewise.
21920         (tablejump): Likewise.
21921         (*sibcall_memory): Likewise.
21922         (*sibcall_value_memory): Likewise.
21923         Peepholes of indirect call and jump via memory: Likewise.
21924         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
21925         (*sibcall_value_GOT_32): Likewise.
21926         * config/i386/predicates.md (indirect_branch_operand): Likewise.
21927         (GOT_memory_operand): Likewise.
21928         (call_insn_operand): Likewise.
21929         (sibcall_insn_operand): Likewise.
21930         (GOT32_symbol_operand): Likewise.
21931         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
21933 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
21935         PR rtl-optimization/83496
21936         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
21937         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
21938         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
21939         redundant insn, if any.
21940         (relax_delay_slots): Likewise.
21941         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
21943 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
21945         PR tree-optimization/83965
21946         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
21947         that grouped statements are part of a reduction chain.  Return
21948         true if the statement is not marked as a reduction itself but
21949         is part of a group.
21950         (vect_recog_dot_prod_pattern): Don't check whether the statement
21951         is part of a group here.
21952         (vect_recog_sad_pattern): Likewise.
21953         (vect_recog_widen_sum_pattern): Likewise.
21955 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
21957         PR debug/84545
21958         * final.c (rest_of_clean_state): Also look for calls inside sequences.
21960 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
21962         PR target/84530
21963         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
21964         the bool argument.
21965         (ix86_output_indirect_function_return): New prototype.
21966         (ix86_split_simple_return_pop_internal): Likewise.
21967         * config/i386/i386.c (indirect_return_via_cx): New.
21968         (indirect_return_via_cx_bnd): Likewise.
21969         (indirect_thunk_name): Handle return va CX_REG.
21970         (output_indirect_thunk_function): Create alias for
21971         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
21972         (ix86_output_indirect_jmp): Remove the bool argument.
21973         (ix86_output_indirect_function_return): New function.
21974         (ix86_split_simple_return_pop_internal): Likewise.
21975         * config/i386/i386.md (*indirect_jump): Don't pass false
21976         to ix86_output_indirect_jmp.
21977         (*tablejump_1): Likewise.
21978         (simple_return_pop_internal): Change it to define_insn_and_split.
21979         Call ix86_split_simple_return_pop_internal to split it for
21980         -mfunction-return=.
21981         (simple_return_indirect_internal): Call
21982         ix86_output_indirect_function_return instead of
21983         ix86_output_indirect_jmp.
21985 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
21987         PR bootstrap/84405
21988         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
21989         memset and value initialization afterwards.
21991 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
21993         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
21995 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
21997         PR target/84521
21998         * common/config/aarch64/aarch64-common.c
21999         (aarch_option_optimization_table[]): Switch
22000         off fomit-frame-pointer
22002 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
22003             Chung-Ju Wu  <jasonwucj@gmail.com>
22005         * config/nds32/nds32-multiple.md (load_multiple): Disallow
22006         volatile memory.
22007         (store_multiple): Ditto.
22009 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
22011         * config.gcc: Add --with-cpu support for nds32 target.
22012         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
22013         * config/nds32/nds32.opt: Add -mcpu= option.
22015 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
22017         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
22018         isel=yes): Warn for these deprecated options.
22020 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
22022         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
22023         ISA_2_5_MASKS_EMBEDDED.
22025 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
22027         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
22028         p->max as pointers rather than using iterative_hash_expr.
22030 2018-02-23  Carl Love  <cel@us.ibm.com>
22032         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
22033         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
22034         BU_P8V_OVERLOAD_2.
22035         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
22036         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
22037         P8V_BUILTIN_VEC_VUNSIGNED2.
22039 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
22041         PR target/81572
22042         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
22043         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
22044         LRA_UNKNOWN_ALT.
22045         * lra-constraints.c (curr_insn_transform): Set up
22046         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
22047         LRA_UNKNOWN_ALT.
22048         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
22049         * lra-eliminations.c (spill_pseudos): Ditto.
22050         (process_insn_for_elimination): Ditto.
22051         * lra-lives.c (reg_early_clobber_p): Use the new macros.
22052         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
22053         LRA_NON_CLOBBERED_ALT.
22055 2018-02-22  Martin Sebor  <msebor@redhat.com>
22057         PR tree-optimization/84480
22058         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
22059         to maybe_diag_stxncpy_trunc.  Call it.
22060         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
22061         from gimple_fold_builtin_strcpy.  Print inlining stack.
22062         (handle_builtin_stxncpy): Print inlining stack.
22063         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
22065 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
22067         PR target/84176
22068         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
22069         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
22070         and -fcheck-pointer-bounds are used together.
22071         (indirect_thunk_prefix): New enum.
22072         (indirect_thunk_need_prefix): New function.
22073         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
22074         "_nt" instead of "_bnd" for NOTRACK prefix.
22075         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
22076         (output_indirect_thunk_function): Likewise.
22077         (): Likewise.
22078         (ix86_code_end): Update output_indirect_thunk_function calls.
22079         (ix86_output_indirect_branch_via_reg): Replace
22080         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
22081         (ix86_output_indirect_branch_via_push): Likewise.
22082         (ix86_output_function_return): Likewise.
22083         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
22084         incompatible with -fcf-protection=branch and
22085         -fcheck-pointer-bounds.
22087 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
22089         PR target/83335
22090         * config/aarch64/aarch64.c (aarch64_print_address_internal):
22091         Change gcc_assert call to output_operand_lossage.
22093 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
22095         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
22097 2018-02-22  DJ Delorie  <dj@redhat.com>
22098             Sebastian Perta  <sebastian.perta@renesas.com>
22099             Oleg Endo  <olegendo@gcc.gnu.org>
22101         * config/rx/rx.c (rx_rtx_costs): New function.
22102         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
22104 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22106         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
22108 2018-02-22  Martin Liska  <mliska@suse.cz>
22110         PR driver/83193
22111         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
22112         Add "native" as a possible value.
22114 2018-02-22  Martin Liska  <mliska@suse.cz>
22116         PR driver/83193
22117         * config/i386/i386.c (ix86_option_override_internal):
22118         Add "native" as a possible value for -march and -mtune.
22120 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
22122         PR target/84502
22123         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
22124         to all type variants.
22126         PR tree-optimization/84503
22127         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
22128         width as info->bitpos + info->bitsize - start.
22129         (merged_store_group::merge_overlapping): Simplify width computation.
22130         (check_no_overlap): New function.
22131         (imm_store_chain_info::try_coalesce_bswap): Compute expected
22132         start + width and last_order of the group, fail if check_no_overlap
22133         fails.
22134         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
22135         to group if check_no_overlap fails.
22137 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
22139         * config/rs6000/altivec.md: Delete contraint arguments to
22140         define_expand, define_split, and define_peephole2, and in
22141         define_insn_and_split if always unused.
22142         * config/rs6000/darwin.md: Ditto.
22143         * config/rs6000/dfp.md: Ditto.
22144         * config/rs6000/rs6000.md: Ditto.
22145         * config/rs6000/sync.md: Ditto.
22146         * config/rs6000/vector.md: Ditto.
22147         * config/rs6000/vsx.md: Ditto.
22149 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
22151         * config/rs6000/altivec.md: Write output control strings as braced
22152         blocks instead of double-quoted strings.
22153         * config/rs6000/darwin.md: Ditto.
22154         * config/rs6000/rs6000.md: Ditto.
22155         * config/rs6000/vector.md: Ditto.
22156         * config/rs6000/vsx.md: Ditto.
22158 2018-02-21  Jason Merrill  <jason@redhat.com>
22160         PR c++/84314 - ICE with templates and fastcall attribute.
22161         * attribs.c (build_type_attribute_qual_variant): Remove assert.
22163 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
22165         * ipa-cp.c (determine_versionability): Fix comment typos.
22167 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
22169         PR c/84229
22170         * ipa-cp.c (determine_versionability): Do not version functions caling
22171         va_arg_pack.
22173 2018-02-21  Martin Liska  <mliska@suse.cz>
22175         PR driver/83193
22176         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
22177         Add "native" as a possible value.
22178         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
22179         the macro when native cpu detection is available.
22181 2018-02-21  Martin Liska  <mliska@suse.cz>
22183         PR driver/83193
22184         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
22185         Add "native" as a possible value.
22186         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
22187         when native cpu detection is available.
22189 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
22190             Martin Sebor  <msebor@redhat.com>
22192         PR tree-optimization/84478
22193         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
22194         false.
22195         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
22196         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
22197         support which is conservatively correct, for 2 only stay conservative
22198         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
22199         argument to the 2 argument get_range_strlen, adjust 6 arg
22200         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
22201         false.
22202         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
22203         (gimple_fold_builtin_strlen): Pass true as last argument to
22204         get_range_strlen.
22206 2018-02-20  Martin Sebor  <msebor@redhat.com>
22208         PR middle-end/84095
22209         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
22210         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
22211         (builtin_memref::builtin_memref): Factor out parts into
22212         set_base_and_offset and call it.
22214 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
22216         PR middle-end/84406
22217         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
22218         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
22219         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
22220         search at the associated MODE_INT.
22222 2018-02-20  Jeff Law  <law@redhat.com>
22224         PR middle-end/82123
22225         PR tree-optimization/81592
22226         PR middle-end/79257
22227         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
22228         for range data rather than using global data.
22229         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
22230         range data rather than using global data.
22231         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
22232         pass it to children as needed.
22233         (struct directive::fmtresult): Similarly.
22234         (struct directive::set_width): Similarly.
22235         (struct directive::set_precision): Similarly.
22236         (format_integer, format_directive, parse_directive): Similarly.
22237         (format_none): Accept unnamed vr_values parameter.
22238         (format_percent, format_floating, format_character): Similarly.
22239         (format_string, format_plain): Similarly.
22240         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
22241         the EVRP range analyzer for range data rather than using global data.
22242         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
22243         gimple-ssa-evrp-analyze.h
22244         (class sprintf_dom_walker): Add after_dom_children member function.
22245         Add evrp_range_analyzer member.
22246         (sprintf_dom_walker::before_dom_children): Call into the EVRP
22247         range analyzer as needed.
22248         (sprintf_dom_walker::after_dom_children): New member function.
22249         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
22250         if not optimizing.
22251         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
22252         (evrp_range_analyzer::pop_to_marker): Likewise.
22254 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
22256         PR tree-optimization/84419
22257         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
22258         with the required type if its current type is compatible but
22259         different.
22261 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
22263         PR middle-end/82004
22264         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
22265         after vectorization.
22267 2018-02-20  Martin Liska  <mliska@suse.cz>
22269         PR driver/83193
22270         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
22271         possible values if we don't have a hint.
22273 2018-02-20  Martin Liska  <mliska@suse.cz>
22275         PR c/84310
22276         PR target/79747
22277         * final.c (shorten_branches): Build align_tab array with one
22278         more element.
22279         * opts.c (finish_options): Add alignment option limit check.
22280         (MAX_CODE_ALIGN): Likewise.
22281         (MAX_CODE_ALIGN_VALUE): Likewise.
22282         * doc/invoke.texi: Document maximum allowed option value for
22283         all -falign-* options.
22285 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
22287         PR target/84146
22288         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
22289         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
22290         * var-tracking.c (emit_note_insn_var_location): Remove all references
22291         to NOTE_INSN_CALL_ARG_LOCATION.
22292         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
22293         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
22294         Use copy_rtx_if_shared.
22295         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
22296         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
22297         (dwarf2out_var_location): Remove handling of
22298         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
22299         on call_insn.
22300         * final.c (final_scan_insn): Remove all references to
22301         NOTE_INSN_CALL_ARG_LOCATION.
22302         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
22303         before dumping final insns.
22304         * except.c (emit_note_eh_region_end): Remove all references to
22305         NOTE_INSN_CALL_ARG_LOCATION.
22306         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
22307         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
22308         * config/arc/arc.c (hwloop_optimize): Likewise.
22309         * config/arm/arm.c (create_fix_barrier): Likewise.
22310         * config/s390/s390.c (s390_chunkify_start): Likewise.
22311         * config/sh/sh.c (find_barrier): Likewise.
22312         * config/i386/i386.c (rest_of_insert_endbranch,
22313         ix86_seh_fixup_eh_fallthru): Likewise.
22314         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
22315         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
22316         * config/frv/frv.c (frv_function_prologue): Likewise.
22317         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
22318         reg note.
22319         (note_outside_basic_block_p): Remove all references to
22320         NOTE_INSN_CALL_ARG_LOCATION.
22321         * gengtype.c (adjust_field_rtx_def): Likewise.
22322         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
22323         Likewise.
22324         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
22325         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
22327         PR c++/84444
22328         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
22329         is ADDR_EXPR.
22331         PR tree-optimization/84452
22332         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
22333         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
22334         is NULL.
22336 2018-02-19  Martin Liska  <mliska@suse.cz>
22338         PR sanitizer/82183
22339         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
22341 2018-02-19  Martin Liska  <mliska@suse.cz>
22342             Richard Sandiford  <richard.sandiford@linaro.org>
22344         PR tree-optimization/82491
22345         * gimple-fold.c (get_base_constructor): Make earlier bail out
22346         to prevent ubsan.
22348 2018-02-19  Carl Love  <cel@us.ibm.com>
22350         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
22351         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
22352         BU_P8V_OVERLOAD_1.
22353         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
22354         P8V_BUILTIN_VEC_NEG.
22356 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
22358         * config/rl78/rl78.md (movdf): New define expand.
22360 2018-02-19  Martin Liska  <mliska@suse.cz>
22362         PR other/80589
22363         * doc/invoke.texi: Fix typo.
22364         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
22366 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
22368         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
22369         handle rs6000_single_float and rs6000_double_float specially for
22370         e500 family CPUs.
22372 2018-02-16  Jeff Law  <law@redhat.com>
22374         * config/rx/rx.c (add_pop_cfi_notes): New function.;
22375         (pop_regs): Use it.
22377 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
22379         PR ipa/84425
22380         * ipa-inline.c (inline_small_functions): Fix a typo.
22382 2018-02-16  Nathan Sidwell  <nathan@acm.org>
22384         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
22386 2018-02-16  Carl Love  <cel@us.ibm.com>
22388         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
22389         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
22390         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
22391         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
22392         expansion to P8V_BUILTIN_VEC_FLOAT2.
22394 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
22396         PR rtl-optimization/70023
22397         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
22398         src_regno into account.
22400 2018-02-16  Carl Love  <cel@us.ibm.com>
22402         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
22403         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
22404         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
22405         * config/rs6000/rs6000.c: Remove case statements for
22406         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
22407         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
22408         and P9V_BUILTIN_VEC_VINSERT4B.
22409         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
22410         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
22411         * config/rs6000/vsx.md:
22412         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
22413         vec_insert4b.
22415 2018-02-16  Carl Love  <cel@us.ibm.com>
22417         * config/rs6000/altivec.h: Add builtin names vec_extract4b
22418         vec_insert4b.
22419         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
22420         definitions.
22421         * config/rs6000/rs6000-c.c: Add the definitions for
22422         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
22423         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
22424         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
22425         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
22426         definition for insert4b and define insn *insert3b_internal.
22427         * doc/extend.texi: Add documentation for vec_extract4b.
22429 2018-02-16  Nathan Sidwell  <nathan@acm.org>
22431         * doc/extend.texi (Backwards Compatibility): Mention friend
22432         injection.  Note for-scope is deprecated.
22433         * doc/invoke.texi (-ffriend-injection): Deprecate.
22435 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
22437         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
22438         that moved to I2, also allow destinations that are a paradoxical
22439         subreg (instead of a normal reg).
22441 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
22443         PR target/83831
22444         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
22445         to QImode.
22447 2018-02-16  Richard Biener  <rguenther@suse.de>
22449         PR tree-optimization/84037
22450         PR tree-optimization/84016
22451         PR target/82862
22452         * config/i386/i386.c (ix86_builtin_vectorization_cost):
22453         Adjust vec_construct for the fact we need additional higher latency
22454         128bit inserts for AVX256 and AVX512 vector builds.
22455         (ix86_add_stmt_cost): Scale vector construction cost for
22456         elementwise loads.
22458 2018-02-16  Richard Biener  <rguenther@suse.de>
22460         PR tree-optimization/84417
22461         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
22462         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
22463         (non_rewritable_lvalue_p): Likewise, use poly-ints.
22465 2018-02-16  Martin Liska  <mliska@suse.cz>
22467         PR sanitizer/84307
22468         * internal-fn.def (ASAN_CHECK): Set proper flags.
22469         (ASAN_MARK): Likewise.
22471 2018-02-16  Julia Koval  <julia.koval@intel.com>
22473         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
22474         from PTA_CANNONLAKE.
22476 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
22478         PR target/84272
22479         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
22480         Use ++iter rather than iter++ for std::list iterators.
22481         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
22482         defer deleting them until all nodes in the forest are processed.  Do
22483         free even leaf nodes.  Change to_process into auto_vec.
22485         PR bootstrap/84405
22486         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
22487         * vec.h (vec_default_construct): Use memset instead of placement new
22488         if BROKEN_VALUE_INITIALIZATION is defined.
22489         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
22490         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
22491         is defined.
22493         PR rtl-optimization/83723
22494         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
22495         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
22496         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
22497         recursive calls.
22498         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
22499         callers.
22500         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
22502 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
22504         PR rtl-optimization/81443
22505         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
22506         from inner REGs to paradoxical SUBREGs.
22508 2018-02-16  Richard Biener  <rguenther@suse.de>
22510         PR tree-optimization/84399
22511         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
22512         For operands we can analyze at their definition make sure we can
22513         analyze them at each use as well.
22515 2018-02-16  Richard Biener  <rguenther@suse.de>
22517         PR tree-optimization/84190
22518         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
22519         volatile accesses if the decl isn't volatile.
22521 2018-02-15  Jason Merrill  <jason@redhat.com>
22523         PR c++/84314 - ICE with templates and fastcall attribute.
22524         * attribs.c (build_type_attribute_qual_variant): Don't clobber
22525         TYPE_CANONICAL on an existing type.
22527 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
22529         PR tree-optimization/84383
22530         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
22531         dstoff nor call operand_equal_p if dstbase is NULL.
22533         PR tree-optimization/84334
22534         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
22535         also a CONSTANT_CLASS_P, punt.
22537 2018-02-14  Jim Wilson  <jimw@sifive.com>
22539         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
22540         first SMALL_OPERAND check.  New local min_second_step.  Move assert
22541         to where locals are set.  Add TARGET_RVC support.
22542         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
22544 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
22546         * doc/invoke.texi: Correct -Wformat-overflow code sample.
22548 2018-02-14  Martin Sebor  <msebor@redhat.com>
22550         PR tree-optimization/83698
22551         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
22552         arrays constrain the offset range to their bounds.
22553         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
22554         (builtin_access::overlap): Avoid setting the size of overlap if it's
22555         already been set.
22556         (maybe_diag_overlap): Also consider arrays when deciding what values
22557         of offsets to include in diagnostics.
22559 2018-02-14  Martin Sebor  <msebor@redhat.com>
22561         PR c/84108
22562         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
22563         that correspond to the kind of a declaration.
22565 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
22567         PR target/83984
22568         * config/pa/pa.md: Load address of PIC label using the linkage table
22569         if the label is nonlocal.
22571 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22573         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
22574         warning message if user requests -maltivec=be.
22575         * doc/invoke.texi: Document deprecation of -maltivec=be.
22577 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
22579         PR target/84220
22580         * config/rs6000/rs6000-c.c: Update definitions for
22581         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
22582         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
22584 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
22586         PR target/84239
22587         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
22588         add _get_ssp intrinsics. Remove argument from
22589         __builtin_ia32_rdssp[d|q].
22590         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
22591         * config/i386/i386-builtin.def: Remove argument from
22592         __builtin_ia32_rdssp[d|q].
22593         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
22594         ix86_expand_special_args_builtin for _rdssp[d|q].
22595         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
22596         Clear register before usage.
22597         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
22598         Add documentation for new _get_ssp and _inc_ssp intrinsics.
22600 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
22602         PR tree-optimization/84357
22603         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
22604         operand 1 of an ARRAY_REF too.
22606 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
22608         PR target/83831
22609         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
22610         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
22611         declarations.
22612         (set_of_reg): New struct.
22613         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
22614         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
22615         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
22616         functions.
22617         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
22618         Split into bitclr, bitset, bitinvert patterns if appropriate.
22619         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
22620         use rx_fuse_in_memory_bitop.
22621         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
22622         to named insn, correct maximum insn length.
22624 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
22626         PR target/79242
22627         * machmode.def: Define a complex mode for PARTIAL_INT.
22628         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
22629         MODE_PARTIAL_INT.
22630         * doc/rtl.texi: Document CSPImode.
22631         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
22632         handling.
22633         (msp430_hard_regno_nregs_with_padding): Likewise.
22635 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
22637         PR target/84279
22638         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
22640 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
22642         PR rtl-optimization/84169
22643         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
22644         we generated a parallel as new i3 and we split that to new i2 and i3
22645         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
22646         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
22647         those to i2, not i1.  Partially rewrite this scan code.
22649 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
22651         PR c/82210
22652         * stor-layout.c (place_field): For variable length fields, adjust
22653         offset_align afterwards not just based on the field's alignment,
22654         but also on the size.
22656         PR middle-end/84309
22657         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
22658         of exps and logs in the use_exp2 case.
22660 2018-02-13  Jeff Law  <law@redhat.com>
22662         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
22663         entry for "vector".
22665         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
22666         ARGS as unused.
22668 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
22670         PR debug/84342
22671         PR debug/84319
22672         * common.opt (gas-loc-support, gas-locview-support): New.
22673         (ginline-points, ginternal-reset-location-views): New.
22674         * doc/invoke.texi: Document them.  Use @itemx where intended.
22675         (gvariable-location-views): Adjust.
22676         * target.def (reset_location_view): New.
22677         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
22678         (TARGET_RESET_LOCATION_VIEW): New.
22679         * doc/tm.texi: Rebuilt.
22680         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
22681         (dwarf2out_default_as_locview_support): New.
22682         (output_asm_line_debug_info): Use option variables.
22683         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
22684         (output_loc_list): Likewise.
22685         (add_high_low_attributes): Check option variables.
22686         Don't output entry view attribute in strict mode.
22687         (gen_inlined_subroutine_die): Check option variables.
22688         (dwarf2out_inline_entry): Likewise.
22689         (init_sections_and_labels): Likewise.
22690         (dwarf2out_early_finish): Likewise.
22691         (maybe_reset_location_view): New, from...
22692         (dwarf2out_var_location): ... here.  Call it.
22693         * debug.h (dwarf2out_default_as_loc_support): Declare.
22694         (dwarf2out_default_as_locview_support): Declare.
22695         * hooks.c (hook_int_rtx_insn_0): New.
22696         * hooks.h (hook_int_rtx_insn_0): Declare.
22697         * toplev.c (process_options): Take -gas-loc-support and
22698         -gas-locview-support from dwarf2out.  Enable
22699         -gvariable-location-views by default only with locview
22700         assembler support.  Enable -ginternal-reset-location-views by
22701         default only if the target defines the corresponding hook.
22702         Enable -ginline-points by default if location views are
22703         enabled; force it disabled if statement frontiers are
22704         disabled.
22705         * tree-inline.c (expand_call_inline): Check option variables.
22706         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
22708 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
22710         PR tree-optimization/84321
22711         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
22712         handling.  Also check whether the anti-range contains any values
22713         that satisfy the mask; switch to a VR_RANGE if not.
22715 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
22717         PR sanitizer/84340
22718         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
22720 2018-02-13  Martin Jambor  <mjambor@suse.cz>
22722         PR c++/83990
22723         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
22724         of call statements, also set location of a load to a temporary.
22726 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
22728         * config/rl78/rl78.c (add_vector_labels): New function.
22729         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
22730         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
22731         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
22732         which checks that no arguments are passed.
22733         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
22734         * doc/extend.texi: Documentation for the new attribute.
22736 2018-02-13  Andreas Schwab  <schwab@suse.de>
22738         * config/riscv/linux.h (CPP_SPEC): Define.
22740 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
22742         PR target/84335
22743         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
22744         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
22745         OPTION_MASK_ISA_AES as first argument to def_builtin_const
22746         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
22747         instead of OPTION_MASK_ISA_PCLMUL as first argument to
22748         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
22749         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
22750         temporarily for AES and PCLMUL builtins.
22752         PR tree-optimization/84339
22753         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
22754         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
22755         Formatting fixes.
22757         PR middle-end/84309
22758         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
22759         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
22760         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
22761         inline function.
22762         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
22763         inline function.
22764         * omp-simd-clone.h: New file.
22765         * omp-simd-clone.c: Include omp-simd-clone.h.
22766         (expand_simd_clones): No longer static.
22767         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
22768         cgraph.h and omp-simd-clone.h.
22769         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
22770         (vect_recog_widen_shift_pattern): Formatting fix.
22771         (vect_pattern_recog_1): Don't check optab for calls.
22773         PR target/84336
22774         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
22775         operands[2] into a REG before using gen_lowpart on it.
22777 2018-02-12  Jeff Law  <law@redhat.com>
22779         PR target/83760
22780         * config/sh/sh.c (find_barrier): Consider a sibling call
22781         a barrier as well.
22783         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
22784         successfully back substituting a reg.
22786 2018-02-12  Richard Biener  <rguenther@suse.de>
22788         PR tree-optimization/84037
22789         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
22790         parameter, move visited init to caller.
22791         (vect_slp_analyze_operations): Separate cost from validity
22792         check, initialize visited once for all instances.
22793         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
22794         for all instances.
22795         * tree-vect-stmts.c (vect_model_simple_cost): Make early
22796         out an assert.
22797         (vect_model_promotion_demotion_cost): Likewise.
22798         (vectorizable_bswap): Guard cost modeling with !slp_node
22799         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
22800         SLP stmts.
22801         (vectorizable_call): Likewise.
22802         (vectorizable_conversion): Likewise.
22803         (vectorizable_assignment): Likewise.
22804         (vectorizable_shift): Likewise.
22805         (vectorizable_operation): Likewise.
22806         (vectorizable_store): Likewise.
22807         (vectorizable_load): Likewise.
22808         (vectorizable_condition): Likewise.
22809         (vectorizable_comparison): Likewise.
22811 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
22813         PR sanitizer/84307
22814         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
22815         (ASAN_MARK): Fix fnspec to account for return value, change pointer
22816         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
22818 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
22820         PR middle-end/83665
22821         * params.def (inline-min-speedup): Increase from 8 to 15.
22822         (max-inline-insns-auto): Decrease from 40 to 30.
22823         * ipa-split.c (consider_split): Add some buffer for function to
22824         be considered inlining candidate.
22825         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
22826         default values.
22828 2018-02-12  Richard Biener  <rguenther@suse.de>
22830         PR tree-optimization/84037
22831         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
22832         matched stmts if we cannot swap the non-matched ones.
22834 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
22836         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
22837         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
22838         _mm_maskz_scalef_round_ss): New intrinsics.
22839         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
22840         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
22841         __builtin_ia32_scalefss_round): Remove.
22842         (__builtin_ia32_scalefsd_mask_round,
22843         __builtin_ia32_scalefss_mask_round): New intrinsics.
22844         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
22845         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
22846         ((match_operand:VF_128 2 "<round_nimm_predicate>"
22847         "<round_constraint>")): Changed to ...
22848         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
22849         "<round_scalar_constraint>")): ... this.
22850         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
22851         %0, %1, %2<round_op3>}"): Changed to ...
22852         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
22853         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
22854         %2<round_scalar_mask_op3>}"): ... this.
22855         * config/i386/subst.md (round_scalar_nimm_predicate): New.
22857 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
22859         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
22860         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
22861         (_mm_maskz_sqrt_round_ss): New intrinsics.
22862         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
22863         (__builtin_ia32_sqrtsd_mask_round)
22864         (__builtin_ia32_sqrtss_mask_round): New builtins.
22865         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
22866         (__builtin_ia32_sqrtss_round): Remove.
22867         (__builtin_ia32_sqrtsd_mask_round)
22868         (__builtin_ia32_sqrtss_mask_round): New builtins.
22869         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
22870         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
22871         ((match_operand:VF_128 1 "vector_operand"
22872         "xBm,<round_constraint>")): Changed to ...
22873         ((match_operand:VF_128 1 "vector_operand"
22874         "xBm,<round_scalar_constraint>")): ... this.
22875         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
22876         %0, %2, %<iptr>1<round_op3>}): Changed to ...
22877         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
22878         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
22879         %<iptr>1<round_scalar_mask_op3>}): ... this.
22880         ((set_attr "prefix" "<round_prefix>")): Changed to ...
22881         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
22883 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
22885         PR target/84266
22886         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
22887         Cast vec_cmpeq result to correct type.
22888         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
22889         Cast vec_cmpgt result to correct type.
22891 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
22893         * final.c (final_scan_insn_1): Renamed from...
22894         (final_scan_insn): ... this.  New wrapper, to recover
22895         seen from the outermost call in recursive ones.
22896         * config/sparc/sparc.c (output_return): Drop seen from call.
22897         (output_sibcall): Likewise.
22898         * config/visium/visium.c (output_branch): Likewise.
22900 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
22902         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
22903         function label.
22905 2018-02-10  Alan Modra  <amodra@gmail.com>
22907         PR target/84300
22908         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
22909         Specify LR as an input.
22911 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
22913         PR sanitizer/83987
22914         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
22915         remove_member_access_dummy_vars): New functions.
22916         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
22917         lower_omp_1, execute_lower_omp): Use them.
22919         PR rtl-optimization/84308
22920         * shrink-wrap.c (spread_components): Release todo vector.
22922 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
22924         PR rtl-optimization/57193
22925         * ira-color.c (struct allocno_color_data): Add member
22926         conflict_allocno_hard_prefs.
22927         (update_conflict_allocno_hard_prefs): New.
22928         (bucket_allocno_compare_func): Add a preference based on
22929         conflict_allocno_hard_prefs.
22930         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
22931         (color_allocnos): Remove a dead code.  Initiate
22932         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
22934 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
22936         PR target/84226
22937         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
22938         constraint from =wa to wa.  Avoid a subreg on the output operand,
22939         instead use a pseudo and subreg it in a move.
22940         (p9_xxbrd_<mode>): Changed to ...
22941         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
22942         (p9_xxbrd_v2df): New expander.
22943         (p9_xxbrw_<mode>): Changed to ...
22944         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
22945         (p9_xxbrw_v4sf): New expander.
22947 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
22949         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
22951 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
22953         PR target/83926
22954         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
22955         multiply in 32-bit mode.
22956         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
22957         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
22958         mode.
22960 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
22962         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
22963         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
22964         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
22965         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
22967 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
22969         PR lto/84213
22970         * dwarf2out.c (is_trivial_indirect_ref): New function.
22971         (dwarf2out_late_global_decl): Do not generate a location
22972         attribute for variables that have a non-trivial DECL_VALUE_EXPR
22973         and that are not defined in the current unit.
22975 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
22977         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
22978         instead of a libcall for UNORDERED.
22980 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
22982         PR target/82641
22983         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
22984         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
22986 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22988         PR target/PR84295
22989         * config/s390/s390.c (s390_set_current_function): Invoke
22990         s390_indirect_branch_settings also if fndecl didn't change.
22992 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
22994         * config/rs6000/rs6000.md (blockage): Set length to zero.
22996 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
22998         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
23000 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
23002         PR sanitizer/84285
23003         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
23004         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
23005         -static-lib*san.
23007         PR debug/84252
23008         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
23009         PARALLEL incoming that failed vt_get_decl_and_offset check.
23011         PR middle-end/84237
23012         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
23013         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
23014         TREE_READONLY bit.
23015         (get_variable_section): For decls in named .bss* sections pass true as
23016         second argument to bss_initializer_p.
23018 2018-02-09  Marek Polacek  <polacek@redhat.com>
23019             Jakub Jelinek  <jakub@redhat.com>
23021         PR c++/83659
23022         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
23023         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
23024         Sync some changes from cxx_fold_indirect_ref.
23026 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
23028         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
23029         markers.
23030         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
23031         (BLOCK_INLINE_ENTRY_LABEL): New.
23032         (dwarf2out_var_location): Disregard inline entry markers.
23033         (inline_entry_data): New struct.
23034         (inline_entry_data_hasher): New hashtable type.
23035         (inline_entry_data_hasher::hash): New.
23036         (inline_entry_data_hasher::equal): New.
23037         (inline_entry_data_table): New variable.
23038         (add_high_low_attributes): Add DW_AT_entry_pc and
23039         DW_AT_GNU_entry_view attributes if a pending entry is found
23040         in inline_entry_data_table.  Add old entry_pc attribute only
23041         if debug nonbinding markers are disabled.
23042         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
23043         markers are enabled.
23044         (block_within_block_p, dwarf2out_inline_entry): New.
23045         (dwarf2out_finish): Check that no entries remained in
23046         inline_entry_data_table.
23047         * final.c (reemit_insn_block_notes): Handle inline entry notes.
23048         (final_scan_insn, notice_source_line): Likewise.
23049         (rest_of_clean_state): Skip inline entry markers.
23050         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
23051         markers.
23052         * gimple.c (gimple_build_debug_inline_entry): New.
23053         * gimple.h (enum gimple_debug_subcode): Add
23054         GIMPLE_DEBUG_INLINE_ENTRY.
23055         (gimple_build_debug_inline_entry): Declare.
23056         (gimple_debug_inline_entry_p): New.
23057         (gimple_debug_nonbind_marker_p): Adjust.
23058         * insn-notes.def (INLINE_ENTRY): New.
23059         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
23060         inline entry marker notes.
23061         (print_insn): Likewise.
23062         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
23063         (INSN_DEBUG_MARKER_KIND): Likewise.
23064         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
23065         * tree-inline.c (expand_call_inline): Build and insert
23066         debug_inline_entry stmt.
23067         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
23068         inline entry blocks early, if nonbind markers are enabled.
23069         (dump_scope_block): Dump fragment info.
23070         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
23071         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
23072         (gimple_build_debug_inline_entry): New.
23073         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
23074         Enable/disable inline entry points too.
23075         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
23076         (DEBUG_INSN): Describe inline entry markers.
23078         * common.opt (gvariable-location-views): New.
23079         (gvariable-location-views=incompat5): New.
23080         * config.in: Rebuilt.
23081         * configure: Rebuilt.
23082         * configure.ac: Test assembler for view support.
23083         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
23084         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
23085         * dwarf2out.c (var_loc_view): New typedef.
23086         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
23087         (dwarf2out_locviews_in_attribute): New.
23088         (dwarf2out_locviews_in_loclist): New.
23089         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
23090         (enum dw_line_info_opcode): Add LI_adv_address.
23091         (struct dw_line_info_table): Add view.
23092         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
23093         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
23094         (zero_view_p): New variable.
23095         (ZERO_VIEW_P): New macro.
23096         (output_asm_line_debug_info): New.
23097         (struct var_loc_node): Add view.
23098         (add_AT_view_list, AT_loc_list): New.
23099         (add_var_loc_to_decl): Add view param.  Test it against last.
23100         (new_loc_list): Add view params.  Record them.
23101         (AT_loc_list_ptr): Handle loc and view lists.
23102         (view_list_to_loc_list_val_node): New.
23103         (print_dw_val): Handle dw_val_class_view_list.
23104         (size_of_die): Likewise.
23105         (value_format): Likewise.
23106         (loc_list_has_views): New.
23107         (gen_llsym): Set vl_symbol too.
23108         (maybe_gen_llsym, skip_loc_list_entry): New.
23109         (dwarf2out_maybe_output_loclist_view_pair): New.
23110         (output_loc_list): Output view list or entries too.
23111         (output_view_list_offset): New.
23112         (output_die): Handle dw_val_class_view_list.
23113         (output_dwarf_version): New.
23114         (output_compilation_unit_header): Use it.
23115         (output_skeleton_debug_sections): Likewise.
23116         (output_rnglists, output_line_info): Likewise.
23117         (output_pubnames, output_aranges): Update version comments.
23118         (output_one_line_info_table): Output view numbers in asm comments.
23119         (dw_loc_list): Determine current endview, pass it to new_loc_list.
23120         Call maybe_gen_llsym.
23121         (loc_list_from_tree_1): Adjust.
23122         (add_AT_location_description): Create view list attribute if
23123         needed, check it's absent otherwise.
23124         (convert_cfa_to_fb_loc_list): Adjust.
23125         (maybe_emit_file): Call output_asm_line_debug_info for test.
23126         (dwarf2out_var_location): Reset views as needed.  Precompute
23127         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
23128         attribute.  Set view.
23129         (new_line_info_table): Reset next view.
23130         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
23131         (dwarf2out_source_line): Likewise.  Output view resets and labels to
23132         the assembler, or select appropriate line info opcodes.
23133         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
23134         (optimize_string_length): Catch it.  Adjust.
23135         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
23136         dw_val_class_view_list, and remove it if no longer needed.
23137         (hash_loc_list): Hash view numbers.
23138         (loc_list_hasher::equal): Compare them.
23139         (optimize_location_lists): Check whether a view list symbol is
23140         needed, and whether the locview attribute is present, and
23141         whether they match.  Remove the locview attribute if no longer
23142         needed.
23143         (index_location_lists): Call skip_loc_list_entry for test.
23144         (dwarf2out_finish): Call output_asm_line_debug_info for test.
23145         Use output_dwarf_version.
23146         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
23147         (struct dw_val_node): Add val_view_list.
23148         * final.c (SEEN_NEXT_VIEW): New.
23149         (set_next_view_needed): New.
23150         (clear_next_view_needed): New.
23151         (maybe_output_next_view): New.
23152         (final_start_function): Rename to...
23153         (final_start_function_1): ... this.  Take pointer to FIRST,
23154         add SEEN parameter.  Emit param bindings in the initial view.
23155         (final_start_function): Reintroduce SEEN-less interface.
23156         (final): Rename to...
23157         (final_1): ... this.  Take SEEN parameter.  Output final pending
23158         next view at the end.
23159         (final): Reintroduce seen-less interface.
23160         (final_scan_insn): Output pending next view before switching
23161         sections or ending a block.  Mark the next view as needed when
23162         outputting variable locations.  Notify debug backend of section
23163         changes, and of location view changes.
23164         (rest_of_handle_final): Adjust.
23165         * toplev.c (process_options): Autodetect value for debug variable
23166         location views option.  Warn on incompat5 without -gdwarf-5.
23167         * doc/invoke.texi (gvariable-location-views): New.
23168         (gvariable-location-views=incompat5): New.
23169         (gno-variable-location-views): New.
23171 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
23173         PR tree-optimization/84136
23174         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
23175         that the result of find_edge is non-NULL.
23177 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
23179         PR target/83008
23180         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
23181         storing integer register in SImode.  Fix cost of 256 and 512
23182         byte aligned SSE register store.
23184 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
23186         * config/i386/i386.c (ix86_multiplication_cost): Fix
23187         multiplication cost for TARGET_AVX512DQ.
23189 2018-02-08  Marek Polacek  <polacek@redhat.com>
23191         PR tree-optimization/84238
23192         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
23193         get_range_strlen.
23195 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
23197         PR tree-optimization/84265
23198         * tree-vect-stmts.c (vectorizable_store): Don't treat
23199         VMAT_CONTIGUOUS accesses as grouped.
23200         (vectorizable_load): Likewise.
23202 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
23204         PR tree-optimization/81635
23205         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
23206         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
23207         (test_round_for_mask): New functions.
23208         (wide_int_cc_tests): Call test_round_for_mask.
23209         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
23210         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
23211         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
23212         range returned by get_range_info.
23214 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
23216         PR ipa/81360
23217         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
23218         * symtab.c: Include builtins.h
23219         (symtab_node::output_to_lto_symbol_table_p): Move here
23220         from lto-streamer-out.c:output_symbol_p.
23221         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
23222         (output_symbol_p): Move all logic to symtab.c
23223         (produce_symtab): Update.
23225 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23227         * config/s390/s390-opts.h (enum indirect_branch): Define.
23228         * config/s390/s390-protos.h (s390_return_addr_from_memory)
23229         (s390_indirect_branch_via_thunk)
23230         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
23231         (enum s390_indirect_branch_type): Define.
23232         * config/s390/s390.c (struct s390_frame_layout, struct
23233         machine_function): Remove.
23234         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
23235         (indirect_branch_table_label_no, indirect_branch_table_name):
23236         Define variables.
23237         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
23238         (enum s390_indirect_branch_option): Define.
23239         (s390_return_addr_from_memory): New function.
23240         (s390_handle_string_attribute): New function.
23241         (s390_attribute_table): Add new attribute handler.
23242         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
23243         (s390_indirect_branch_via_thunk): New function.
23244         (s390_indirect_branch_via_inline_thunk): New function.
23245         (s390_function_ok_for_sibcall): When jumping via thunk disallow
23246         sibling call optimization for non z10 compiles.
23247         (s390_emit_call): Force indirect branch target to be a single
23248         register.  Add r1 clobber for non-z10 compiles.
23249         (s390_emit_epilogue): Emit return jump via return_use expander.
23250         (s390_reorg): Handle JUMP_INSNs as execute targets.
23251         (s390_option_override_internal): Perform validity checks for the
23252         new command line options.
23253         (s390_indirect_branch_attrvalue): New function.
23254         (s390_indirect_branch_settings): New function.
23255         (s390_set_current_function): Invoke s390_indirect_branch_settings.
23256         (s390_output_indirect_thunk_function):  New function.
23257         (s390_code_end): Implement target hook.
23258         (s390_case_values_threshold): Implement target hook.
23259         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
23260         macros.
23261         * config/s390/s390.h (struct s390_frame_layout)
23262         (struct machine_function): Move here from s390.c.
23263         (TARGET_INDIRECT_BRANCH_NOBP_RET)
23264         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
23265         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
23266         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
23267         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
23268         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
23269         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
23270         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
23271         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
23272         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
23273         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
23274         (mnemonic attribute): Add values which aren't recognized
23275         automatically.
23276         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
23277         pattern for branch conversion.  Fix mnemonic attribute.
23278         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
23279         indirect branch via thunk if requested.
23280         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
23281         ("*indirect_jump"): Disable for branch conversion using out of
23282         line thunks.
23283         ("indirect_jump_via_thunk<mode>_z10")
23284         ("indirect_jump_via_thunk<mode>")
23285         ("indirect_jump_via_inlinethunk<mode>_z10")
23286         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
23287         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
23288         ("casesi_jump_via_inlinethunk<mode>_z10")
23289         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
23290         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
23291         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
23292         ("*indirect2_jump"): Disable for branch conversion.
23293         ("casesi_jump"): Turn into expander and expand patterns for branch
23294         conversion.
23295         ("return_use"): New expander.
23296         ("*return"): Emit return via thunk and rename it to ...
23297         ("*return<mode>"): ... this one.
23298         * config/s390/s390.opt: Add new options and and enum for the
23299         option values.
23301 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
23303         * lra-constraints.c (match_reload): Unconditionally use
23304         gen_lowpart_SUBREG, rather than selecting between that
23305         and equivalent gen_rtx_SUBREG code.
23307 2018-02-08  Richard Biener  <rguenther@suse.de>
23309         PR tree-optimization/84233
23310         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
23311         changed flag instead of boguously re-using phi_inserted.
23313 2018-02-08  Martin Jambor  <mjambor@suse.cz>
23315         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
23316         static local variables.
23318 2018-02-08  Richard Biener  <rguenther@suse.de>
23320         PR tree-optimization/84278
23321         * tree-vect-stmts.c (vectorizable_store): When looking for
23322         smaller vector types to perform grouped strided loads/stores
23323         make sure the mode is supported by the target.
23324         (vectorizable_load): Likewise.
23326 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
23328         * config/aarch64/aarch64.c (aarch64_components_for_bb):
23329         Increase LDP/STP opportunities by adding adjacent callee-saves.
23331 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
23333         PR rtl-optimization/84068
23334         PR rtl-optimization/83459
23335         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
23337 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
23339         PR tree-optimization/84224
23340         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
23341         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
23342         non-zero arguments.
23344 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
23346         PR target/84113
23347         * config/rs6000/altivec.md (*restore_world): Remove LR use.
23348         * config/rs6000/predicates.md (restore_world_operation): Adjust op
23349         count, remove one USE.
23351 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
23353         * doc/install.texi (Configuration): Document the
23354         --with-long-double-format={ibm,ieee} PowerPC configuration
23355         options.
23357         PR target/84154
23358         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
23359         Convert from define_expand to be define_insn_and_split.  Rework
23360         float/double/_Float128 conversions to QI/HI/SImode to work with
23361         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
23362         conversions to QI/HImode types did a store and then a load to
23363         truncate the value.  For conversions to VSX registers, don't split
23364         the insn, instead emit the code directly.  Use the code iterator
23365         any_fix to combine signed and unsigned conversions.
23366         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
23367         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
23368         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
23369         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
23370         (fix_<mode>di2_hw): Likewise.
23371         (fixuns_<mode>di2_hw): Likewise.
23372         (fix_<mode>si2_hw): Likewise.
23373         (fixuns_<mode>si2_hw): Likewise.
23374         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
23375         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
23376         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
23377         fix<uns>_trunc<SFDF:mode>si2_p8.
23378         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
23379         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23380         (fix<uns>_<mode>_mem): Likewise.
23381         (fctiw<u>z_<mode>_mem): Likewise.
23382         (fix<uns>_<mode>_mem): Likewise.
23383         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
23384         the register allocator from doing a direct move to the GPRs to do
23385         a store, and instead use the ISA 3.0 store byte/half-word from
23386         vector register instruction.  For IEEE 128-bit floating point,
23387         also optimize stores of 32-bit ints.
23388         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
23390 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
23392         * genextract.c (push_pathstr_operand): New function to support
23393         [a-zA-Z].
23394         (walk_rtx): Call push_pathstr_operand.
23395         (print_path): Support [a-zA-Z].
23397 2018-02-07  Richard Biener  <rguenther@suse.de>
23399         PR tree-optimization/84037
23400         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
23401         (cse_and_gimplify_to_preheader): Declare.
23402         (vect_get_place_in_interleaving_chain): Likewise.
23403         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
23404         ivexpr_map.
23405         (_loop_vec_info::~_loop_vec_info): Delete it.
23406         (cse_and_gimplify_to_preheader): New function.
23407         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
23408         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
23409         (vectorizable_load): Likewise.  For grouped stores always base
23410         the IV on the first element.
23411         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
23412         condition before gimplifying.
23414 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
23416         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
23417         *DIV_EXPR and *MOD_EXPR.
23419 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
23421         PR target/84248
23422         * config/i386/i386.c (ix86_option_override_internal): Mask out
23423         the CF_SET bit when checking -fcf-protection.
23425 2018-02-07  Tom de Vries  <tom@codesourcery.com>
23427         PR libgomp/84217
23428         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
23429         enough.
23431 2018-02-07  Richard Biener  <rguenther@suse.de>
23433         PR tree-optimization/84204
23434         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
23435         this place.
23437         PR tree-optimization/84205
23438         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
23439         special-case isl_ast_op_zdiv_r.
23441         PR tree-optimization/84223
23442         * graphite-scop-detection.c (gather_bbs::before_dom_children):
23443         Only add conditions from within the region.
23444         (gather_bbs::after_dom_children): Adjust.
23446 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
23448         PR target/84209
23449         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
23450         * config/avr/avr.md: Only post-reload split REG-REG moves if
23451         either register is GENERAL_REG_P.
23453 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
23455         PR tree-optimization/84235
23456         * tree-ssa-scopedtables.c
23457         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
23458         if the subtraction is performed in floating point type where NaNs are
23459         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
23460         build 1.  Formatting fix.
23462 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
23464         PR target/84146
23465         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
23466         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
23467         and skip it regardless of bb boundaries.  Use CALL_P macro,
23468         don't test INSN_P (insn) together with CALL_P or JUMP_P check
23469         unnecessarily, formatting fix.
23471 2018-02-06  Michael Collison  <michael.collison@arm.com>
23473         * config/arm/thumb2.md:
23474         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
23475         (*thumb_mov_notscc): Ditto.
23477 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
23479         PR target/84154
23480         * config/rs6000/rs6000.md (su code attribute): Use "u" for
23481         unsigned_fix, not "s".
23483 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23485         * configure.ac (gcc_fn_eh_frame_ro): New function.
23486         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
23487         correct .eh_frame permissions.
23488         * configure: Regenerate.
23490 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
23492         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
23493         irrelevant options.
23495 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23497         * config/rs6000/rs6000.c (rs6000_option_override_internal):
23498         Display warning message for -mno-speculate-indirect-jumps.
23500 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
23502         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
23503         Undocumented.
23504         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
23506 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
23508         PR tree-optimization/84225
23509         * tree-eh.c (find_trapping_overflow): Only call
23510         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
23512 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
23514         PR target/84145
23515         * config/i386/i386.c: Reimplement the check of possible options
23516         -mibt/-mshstk conbination. Change error messages.
23517         * doc/invoke.texi: Fix a typo: remove extra '='.
23519 2018-02-06  Marek Polacek  <polacek@redhat.com>
23521         PR tree-optimization/84228
23522         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
23524 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
23526         PR target/82641
23527         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
23528         emitted arch directives.
23529         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
23530         __ARM_FEATURE_COPROC before changing architectures.
23532 2018-02-06  Richard Biener  <rguenther@suse.de>
23534         * config/i386/i386.c (print_reg): Fix typo.
23535         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
23537 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
23539         * configure: Regenerate.
23541 2018-02-05  Martin Sebor  <msebor@redhat.com>
23543         PR tree-optimization/83369
23544         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
23545         inlining context.
23547 2018-02-05  Martin Liska  <mliska@suse.cz>
23549         * doc/invoke.texi: Cherry-pick upstream r323995.
23551 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
23553         * ira.c (ira_init_register_move_cost): Adjust comment.
23555 2018-02-05  Martin Liska  <mliska@suse.cz>
23557         PR gcov-profile/84137
23558         * doc/gcov.texi: Fix typo in documentation.
23560 2018-02-05  Martin Liska  <mliska@suse.cz>
23562         PR gcov-profile/83879
23563         * doc/gcov.texi: Document necessity of --dynamic-list-data when
23564         using dlopen functionality.
23566 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
23568         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
23569         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
23570         _mm_maskz_range_ss, _mm_mask_range_round_ss,
23571         _mm_maskz_range_round_ss): New intrinsics.
23572         (__builtin_ia32_rangesd128_round)
23573         (__builtin_ia32_rangess128_round): Remove.
23574         (__builtin_ia32_rangesd128_mask_round,
23575         __builtin_ia32_rangess128_mask_round): New builtins.
23576         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
23577         __builtin_ia32_rangess128_round): Remove.
23578         (__builtin_ia32_rangesd128_mask_round,
23579         __builtin_ia32_rangess128_mask_round): New builtins.
23580         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
23581         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
23582         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
23583         "<round_saeonly_constraint>")): Changed to ...
23584         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
23585         "<round_saeonly_scalar_constraint>")): ... this.
23586         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
23587         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
23588         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
23589         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
23590         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
23592 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
23594         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
23595         options.
23596         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
23597         Remove all values except native, 8540 and 8548.
23599 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
23601         * config/i386/i386.c (ix86_output_function_return): Pass
23602         INVALID_REGNUM, instead of -1, as invalid register number to
23603         indirect_thunk_name and output_indirect_thunk.
23605 2018-02-02  Julia Koval  <julia.koval@intel.com>
23607         * config.gcc: Add -march=icelake.
23608         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
23609         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
23610         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
23611         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
23612         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
23613         (processor_target_table): Add icelake.
23614         (ix86_option_override_internal): Handle new PTAs.
23615         (get_builtin_code_for_version): Handle icelake.
23616         (M_INTEL_COREI7_ICELAKE): New.
23617         (fold_builtin_cpu): Handle icelake.
23618         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
23619         * doc/invoke.texi: Add -march=icelake.
23621 2018-02-02  Julia Koval  <julia.koval@intel.com>
23623         * config/i386/i386.c (ix86_option_override_internal): Change flags type
23624         to wide_int_bitmask.
23625         * wide-int-bitmask.h: New.
23627 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
23629         PR target/84066
23630         * config/i386/i386.md: Replace Pmode with word_mode in
23631         builtin_setjmp_setup and builtin_longjmp to support x32.
23633 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
23635         PR target/56010
23636         PR target/83743
23637         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
23638         #include "opts.h".
23639         (rs6000_supported_cpu_names): New static variable.
23640         (linux_cpu_translation_table): Likewise.
23641         (elf_platform) <cpu>: Define new static variable and use it.
23642         Translate kernel AT_PLATFORM name to canonical name if needed.
23643         Error if platform name is unknown.
23645 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
23647         PR target/84089
23648         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
23650 2018-02-01  Jeff Law  <law@redhat.com>
23652         PR target/84128
23653         * config/i386/i386.c (release_scratch_register_on_entry): Add new
23654         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
23655         the scratch if RELEASE_VIA_POP is false.
23656         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
23657         If we have to save a temporary register, decrement SIZE appropriately.
23658         Pass new arguments to release_scratch_register_on_entry.
23659         (ix86_adjust_stack_and_probe): Likewise.
23660         (ix86_emit_probe_stack_range): Pass new arguments to
23661         release_scratch_register_on_entry.
23663 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
23665         PR rtl-optimization/84157
23666         * combine.c (change_zero_ext): Use REG_P predicate in
23667         front of HARD_REGISTER_P predicate.
23669 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
23671         * config/avr/avr.c (avr_option_override): Move disabling of
23672         -fdelete-null-pointer-checks to...
23673         * common/config/avr/avr-common.c (avr_option_optimization_table):
23674         ...here.
23676 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23678         PR tree-optimization/81635
23679         * tree-data-ref.c (split_constant_offset_1): For types that
23680         wrap on overflow, try to use range info to prove that wrapping
23681         cannot occur.
23683 2018-02-01  Renlin Li  <renlin.li@arm.com>
23685         PR target/83370
23686         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
23687         TAILCALL_ADDR_REGS.
23688         (aarch64_register_move_cost): Likewise.
23689         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
23690         TAILCALL_ADDR_REGS.
23691         (REG_CLASS_NAMES): Likewise.
23692         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
23693         TAILCALL_ADDR_REGS. Remove IP registers.
23694         * config/aarch64/aarch64.md (Ucs): Update register constraint.
23696 2018-02-01  Richard Biener  <rguenther@suse.de>
23698         * domwalk.h (dom_walker::dom_walker): Add additional constructor
23699         for specifying RPO order and allow NULL for that.
23700         * domwalk.c (dom_walker::dom_walker): Likewise.
23701         (dom_walker::walk): Handle NULL RPO order.
23702         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
23703         in RPO order.
23704         (rewrite_update_dom_walker): Likewise.
23705         (mark_def_dom_walker): Likewise.
23707 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23709         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
23710         (aarch64_maybe_expand_sve_subreg_move): Declare.
23711         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
23712         * config/aarch64/predicates.md (aarch64_any_register_operand): New
23713         predicate.
23714         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
23715         that are semantically a reverse operation.
23716         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
23717         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
23718         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
23719         functions.
23720         (aarch64_can_change_mode_class): For big-endian, forbid changes
23721         between two SVE modes if they have different element sizes.
23723 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23725         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
23726         the TImode handling for big-endian targets.
23728 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23730         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
23731         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
23732         not just bytes.
23733         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
23734         Remove BSWAP handing for big-endian targets and use the form of
23735         LD1RQ appropariate for the mode.
23737 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23739         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
23740         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
23741         duplicated element.
23743 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
23745         PR tearget/83845
23746         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
23747         check for operands that need to go through aarch64_sve_reload_be.
23749 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
23751         PR tree-optimization/81661
23752         PR tree-optimization/84117
23753         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
23754         * tree-eh.c: Include gimplify.h.
23755         (find_trapping_overflow, replace_trapping_overflow,
23756         rewrite_to_non_trapping_overflow): New functions.
23757         * tree-vect-loop.c: Include tree-eh.h.
23758         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
23759         * tree-data-ref.c: Include tree-eh.h.
23760         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
23762 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
23764         PR rtl-optimization/84123
23765         * combine.c (change_zero_ext): Check if hard register satisfies
23766         can_change_dest_mode before calling gen_lowpart_SUBREG.
23768 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
23770         PR target/82444
23771         * ira.c (ira_init_register_move_cost): Remove assert.
23773 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
23775         PR rtl-optimization/84071
23776         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
23777         * doc/tm.texi: Regenerate.
23779 2018-01-31  Richard Biener  <rguenther@suse.de>
23781         PR tree-optimization/84132
23782         * tree-data-ref.c (analyze_miv_subscript): Properly
23783         check whether evolution_function_is_affine_multivariate_p
23784         before calling gcd_of_steps_may_divide_p.
23786 2018-01-31  Julia Koval  <julia.koval@intel.com>
23788         PR target/83618
23789         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
23790         * config/i386/i386.md (rdpid_rex64) New.
23791         (rdpid): Make 32bit only.
23793 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
23795         PR lto/84105
23796         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
23797         an IDENTIFIER_NODE for FUNCTION_TYPE's.
23799 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
23801         Revert
23802         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
23804         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
23806 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
23808         PR rtl-optimization/84071
23809         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
23810         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
23812 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
23814         * config/arc/arc.c (arc_handle_aux_attribute): New function.
23815         (arc_attribute_table): Add 'aux' attribute.
23816         (arc_in_small_data_p): Consider aux like variables.
23817         (arc_is_aux_reg_p): New function.
23818         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
23819         (arc_get_aux_arg): New function.
23820         (prepare_move_operands): Handle aux-register access.
23821         (arc_handle_aux_attribute): New function.
23822         * doc/extend.texi (ARC Variable attributes): Add subsection.
23824 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
23826         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
23827         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
23828         (arc_attribute_table): Add 'uncached' attribute.
23829         (arc_print_operand): Print '.di' flag for uncached memory
23830         accesses.
23831         (arc_in_small_data_p): Do not consider for small data the uncached
23832         types.
23833         (arc_is_uncached_mem_p): New function.
23834         * config/arc/predicates.md (compact_store_memory_operand): Check
23835         for uncached memory accesses.
23836         (nonvol_nonimm_operand): Likewise.
23837         * doc/extend.texi (ARC Type Attribute): New subsection.
23839 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
23841         PR c/84100
23842         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
23843         falign-loops=): Add Optimization flag.
23845 2018-01-30  Jeff Law  <law@redhat.com>
23847         PR target/84064
23848         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
23849         INT_REGISTERS_SAVED.  Check it prior to calling
23850         get_scratch_register_on_entry.
23851         (ix86_adjust_stack_and_probe): Similarly.
23852         (ix86_emit_probe_stack_range): Similarly.
23853         (ix86_expand_prologue): Corresponding changes.
23855 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23857         PR target/40411
23858         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
23859         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
23861 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
23863         PR target/84112
23864         * lra-constraints.c (curr_insn_transform): Process AND in the
23865         address.
23867 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
23869         PR rtl-optimization/83986
23870         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
23871         dependence against last_pending_memory_flush in addition to
23872         pending_jump_insns.
23874 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
23876         PR tree-optimization/81611
23877         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
23878         copies.
23880 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23882         PR target/83758
23883         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
23884         a reg rtx.
23886 2018-01-30  Richard Biener  <rguenther@suse.de>
23887             Jakub Jelinek  <jakub@redhat.com>
23889         PR tree-optimization/84111
23890         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
23891         inner loops added during recursion, as they don't have up-to-date
23892         SSA form.
23894 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
23896         PR ipa/81360
23897         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
23898         (can_inline_edge_by_limits_p): ... here.
23899         (can_early_inline_edge_p, check_callers,
23900         update_caller_keys, update_callee_keys, recursive_inlining,
23901         add_new_edges_to_heap, speculation_useful_p,
23902         inline_small_functions,
23903         inline_small_functions, flatten_function,
23904         inline_to_all_callers_1): Update.
23906 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
23908         * profile-count.c (profile_count::combine_with_ipa_count): Handle
23909         zeros correctly.
23911 2018-01-30  Richard Biener  <rguenther@suse.de>
23913         PR tree-optimization/83008
23914         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
23915         invariant and constant vector uses in stmts when they need
23916         more than one stmt.
23918 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23920         PR bootstrap/84017
23921         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
23922         * configure: Regenerate.
23924 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
23926         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
23927         pattern.
23928         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
23929         Use gen_rtx_REG rather than gen_lowpart.
23931 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
23933         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
23934         rather than 0 when creating partial subregs.
23936 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
23938         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
23939         of usage.
23941 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
23943         PR target/81550
23944         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
23945         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
23946         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
23947         flags.  This restores the settings used before the 2017-07-24.
23948         Turning off pre increment/decrement/modify allows IVOPTS to
23949         optimize DF/SF loops where the index is an int.
23951 2018-01-29  Richard Biener  <rguenther@suse.de>
23952             Kelvin Nilsen  <kelvin@gcc.gnu.org>
23954         PR bootstrap/80867
23955         * tree-vect-stmts.c (vectorizable_call): Don't call
23956         targetm.vectorize_builtin_md_vectorized_function if callee is
23957         NULL.
23959 2018-01-22  Carl Love  <cel@us.ibm.com>
23961         * doc/extend.tex: Fix typo in second arg in
23962         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
23964 2018-01-29  Richard Biener  <rguenther@suse.de>
23966         PR tree-optimization/84086
23967         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
23968         (flush_ssaname_freelist): When SSA names were released reset
23969         the SCEV hash table.
23971 2018-01-29  Richard Biener  <rguenther@suse.de>
23973         PR tree-optimization/84057
23974         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
23975         removed paths when removing edges.
23977 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
23979         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
23980         -mfunction-return=@var{choice}.
23982 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
23984         PR diagnostic/84034
23985         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
23986         Handle CR like TAB.
23987         (layout::print_source_line): Likewise.
23988         (test_get_line_width_without_trailing_whitespace): Add test cases.
23990 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
23992         PR middle-end/84040
23993         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
23994         debug insns.
23996 2018-01-26  Jim Wilson  <jimw@sifive.com>
23998         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
24000         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
24001         specified.
24003 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24005         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
24006         and CMP + SUB-immediate -> SUBS.
24008 2018-01-26  Martin Sebor  <msebor@redhat.com>
24010         PR tree-optimization/83896
24011         * tree-ssa-strlen.c (get_string_len): Rename...
24012         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
24013         Avoid assuming length is constant.
24014         (handle_char_store): Use HOST_WIDE_INT for string length.
24016 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
24018         PR target/81763
24019         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
24020         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
24022 2018-01-26  Richard Biener  <rguenther@suse.de>
24024         PR rtl-optimization/84003
24025         * dse.c (record_store): Only record redundant stores when
24026         the earlier store aliases at least all accesses the later one does.
24028 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
24030         PR rtl-optimization/83985
24031         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
24032         REG_CFA_RESTORE insns.
24033         (delete_unmarked_insns): Don't ignore separate shrink wrapping
24034         REG_CFA_RESTORE insns here.
24036         PR c/83989
24037         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
24038         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
24040 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24042         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
24043         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
24044         (arc_init): Likewise.
24045         (arc_override_options): Likewise.
24046         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
24047         value.
24048         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
24049         support.
24050         * config/arc/arc.h (TARGET_DBNZ): Define.
24051         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
24052         properly set the tune attribute.
24053         (dbnz): Use TARGET_DBNZ guard.
24054         * config/arc/arc.opt (mtune): Add core3 option.
24056 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24058         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
24059         recognize new pic like addresses.
24060         (arc_delegitimize_address): Clean up.
24062 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24064         * config/arc/arc-arches.def: Option mrf16 valid for all
24065         architectures.
24066         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
24067         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
24068         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
24069         * config/arc/arc-tables.opt: Regenerate.
24070         * config/arc/arc.c (arc_conditional_register_usage): Handle
24071         reduced register file case.
24072         (arc_file_start): Set must have build attributes.
24073         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
24074         mrf16 option value.
24075         * config/arc/arc.opt (mrf16): Add new option.
24076         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
24077         * config/arc/genmultilib.awk: Handle new mrf16 option.
24078         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
24079         * config/arc/t-multilib: Regenerate.
24080         * doc/invoke.texi (ARC Options): Document mrf16 option.
24082 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24084         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
24085         * config/arc/arc.c (arc_handle_secure_attribute): New function.
24086         (arc_attribute_table): Add 'secure_call' attribute.
24087         (arc_print_operand): Print secure call operand.
24088         (arc_function_ok_for_sibcall): Don't optimize tail calls when
24089         secure.
24090         (arc_is_secure_call_p): New function.  * config/arc/arc.md
24091         (call_i): Add support for sjli instruction.
24092         (call_value_i): Likewise.
24093         * config/arc/constraints.md (Csc): New constraint.
24095 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
24096             John Eric Martin  <John.Martin@emmicro-us.com>
24098         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
24099         * config/arc/arc.c (_arc_jli_section): New struct.
24100         (arc_jli_section): New type.
24101         (rc_jli_sections): New static variable.
24102         (arc_handle_jli_attribute): New function.
24103         (arc_attribute_table): Add jli_always and jli_fixed attribute.
24104         (arc_file_end): New function.
24105         (TARGET_ASM_FILE_END): Define.
24106         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
24107         (arc_add_jli_section): New function.
24108         (jli_call_scan): Likewise.
24109         (arc_reorg): Call jli_call_scan.
24110         (arc_output_addsi): Remove 'S' from printing asm operand.
24111         (arc_is_jli_call_p): New function.
24112         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
24113         operand.
24114         (movhi_insn): Likewise.
24115         (movsi_insn): Likewise.
24116         (movsi_set_cc_insn): Likewise.
24117         (loadqi_update): Likewise.
24118         (load_zeroextendqisi_update): Likewise.
24119         (load_signextendqisi_update): Likewise.
24120         (loadhi_update): Likewise.
24121         (load_zeroextendhisi_update): Likewise.
24122         (load_signextendhisi_update): Likewise.
24123         (loadsi_update): Likewise.
24124         (loadsf_update): Likewise.
24125         (movsicc_insn): Likewise.
24126         (bset_insn): Likewise.
24127         (bxor_insn): Likewise.
24128         (bclr_insn): Likewise.
24129         (bmsk_insn): Likewise.
24130         (bicsi3_insn): Likewise.
24131         (cmpsi_cc_c_insn): Likewise.
24132         (movsi_ne): Likewise.
24133         (movsi_cond_exec): Likewise.
24134         (clrsbsi2): Likewise.
24135         (norm_f): Likewise.
24136         (normw): Likewise.
24137         (swap): Likewise.
24138         (divaw): Likewise.
24139         (flag): Likewise.
24140         (sr): Likewise.
24141         (kflag): Likewise.
24142         (ffs): Likewise.
24143         (ffs_f): Likewise.
24144         (fls): Likewise.
24145         (call_i): Remove 'S' asm letter, add jli instruction.
24146         (call_value_i): Likewise.
24147         * config/arc/arc.op (mjli-always): New option.
24148         * config/arc/constraints.md (Cji): New constraint.
24149         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
24150         operand.
24151         (subsf3_fpx): Likewise.
24152         (mulsf3_fpx): Likewise.
24153         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
24154         asm operand.
24155         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
24156         function attrbutes.
24157         * doc/invoke.texi (ARC): Document mjli-always option.
24159 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
24161         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
24162         avoid addition with 0 and use incw and decw where possible.
24164 2018-01-26  Richard Biener  <rguenther@suse.de>
24166         PR tree-optimization/81082
24167         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
24168         association if it requires casting to unsigned.
24169         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
24170         from fold_plusminus_mult_expr to catch important cases late when
24171         range info is available.
24173 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24175         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
24176         * configure.ac (hidden_linkonce): New test.
24177         * configure: Regenerate.
24178         * config.in: Regenerate.
24180 2018-01-26  Julia Koval  <julia.koval@intel.com>
24182         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
24183         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
24184         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
24185         _mm_mask_bitshuffle_epi64_mask): Fix type.
24186         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
24187         USI_FTYPE_V4DI_V4DI_USI): Remove.
24188         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
24189         __builtin_ia32_vpshufbitqmb256_mask,
24190         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
24191         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
24192         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
24194 2018-01-26  Alan Modra  <amodra@gmail.com>
24196         PR target/84033
24197         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
24198         UNSPEC_VBPERMQ.  Sort other unspecs.
24200 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
24202         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
24204 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
24206         PR middle-end/83055
24207         * predict.c (drop_profile): Do not push/pop cfun; update also
24208         node->count.
24209         (handle_missing_profiles): Fix logic looking for zero profiles.
24211 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
24213         PR middle-end/83977
24214         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
24215         on functions with #pragma omp declare simd or functions with simd
24216         attribute.
24217         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
24218         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
24219         Remove trailing \n from warning_at calls.
24221 2018-01-25  Tom de Vries  <tom@codesourcery.com>
24223         PR target/84028
24224         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
24225         for neutered workers.
24227 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
24229         PR target/68467
24230         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
24231         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
24233 2018-01-24  Jeff Law  <law@redhat.com>
24235         PR target/83994
24236         * i386.c (get_probe_interval): Move to earlier point.
24237         (ix86_compute_frame_layout): If -fstack-clash-protection and
24238         the frame is larger than the probe interval, then use pushes
24239         to save registers rather than reg->mem moves.
24240         (ix86_expand_prologue): Remove conditional for int_registers_saved
24241         assertion.
24243 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
24245         PR target/84014
24246         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
24247         min/max for never referenced object.
24249 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
24251         PR middle-end/83977
24252         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
24253         here.
24254         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
24255         attributes from DECL_ATTRIBUTES (decl) without affecting
24256         DECL_ATTRIBUTES (current_function_decl).
24257         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
24258         functions with non-NULL DECL_ABSTRACT_ORIGIN.
24260 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
24262         PR tree-optimization/83979
24263         * fold-const.c (fold_comparison): Use constant_boolean_node
24264         instead of boolean_{true,false}_node.
24266 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
24268         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
24269         with zero counts.
24271 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24273         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24274         Simplify the clause that sets the length attribute.
24275         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24276         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
24277         clause that sets the length attribute.
24278         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24280 2018-01-24  Tom de Vries  <tom@codesourcery.com>
24282         PR target/83589
24283         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
24284         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
24285         Add strict parameter.
24286         (prevent_branch_around_nothing): Insert dummy insn between branch to
24287         label and label with no ptx insn inbetween.
24288         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
24290 2018-01-24  Tom de Vries  <tom@codesourcery.com>
24292         PR target/81352
24293         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
24294         for neutered threads in warp.
24295         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
24297 2018-01-24  Richard Biener  <rguenther@suse.de>
24299         PR tree-optimization/83176
24300         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
24301         operands.
24303 2018-01-24  Richard Biener  <rguenther@suse.de>
24305         PR tree-optimization/82819
24306         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
24307         code generating pluses that are no-ops in the target precision.
24309 2018-01-24  Richard Biener  <rguenther@suse.de>
24311         PR middle-end/84000
24312         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
24314 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
24316         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
24317         to merge probabilities.
24318         * predict.c (probably_never_executed): Also mark as cold functions
24319         with global 0 profile and guessed local profile.
24320         * profile-count.c (profile_probability::combine_with_count): New
24321         member function.
24322         * profile-count.h (profile_probability::operator*,
24323         profile_probability::operator*=, profile_probability::operator/,
24324         profile_probability::operator/=): Reduce precision to adjusted
24325         and set value to guessed on contradictory divisions.
24326         (profile_probability::combine_with_freq): Remove.
24327         (profile_probability::combine_wiht_count): Declare.
24328         (profile_count::force_nonzero):: Set to adjusted.
24329         (profile_count::probability_in):: Set quality to adjusted.
24330         * tree-ssa-tail-merge.c (replace_block_by): Use
24331         combine_with_count.
24333 2018-01-23  Andrew Waterman  <andrew@sifive.com>
24334             Jim Wilson  <jimw@sifive.com>
24336         * config/riscv/riscv.c (riscv_stack_boundary): New.
24337         (riscv_option_override): Set riscv_stack_boundary.  Handle
24338         riscv_preferred_stack_boundary_arg.
24339         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
24340         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
24341         (STACK_BOUNDARY): Set to riscv_stack_boundary.
24342         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
24343         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
24344         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
24346 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
24348         PR target/83905
24349         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
24350         of struct ix86_frame.
24351         (ix86_expand_epilogue): Likewise.  Add a local variable for
24352         the reg_save_offset field in struct ix86_frame.
24354 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
24356         PR tree-optimization/82604
24357         * tree-loop-distribution.c (enum partition_kind): New enum item
24358         PKIND_PARTIAL_MEMSET.
24359         (partition_builtin_p): Support above new enum item.
24360         (generate_code_for_partition): Ditto.
24361         (compute_access_range): Differentiate cases that equality can be
24362         proven at all loops, the innermost loops or no loops.
24363         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
24364         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
24365         (finalize_partitions, distribute_loop): Don't fuse partition of
24366         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
24367         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
24368         parloop is enabled.
24370 2018-01-23  Martin Liska  <mliska@suse.cz>
24372         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
24373         order to ignore the predictor.
24374         (PRED_POLYMORPHIC_CALL): Likewise.
24375         (PRED_RECURSIVE_CALL): Likewise.
24377 2018-01-23  Martin Liska  <mliska@suse.cz>
24379         * tree-profile.c (tree_profiling): Print function header to
24380         aware reader which function we are working on.
24381         * value-prof.c (gimple_find_values_to_profile): Do not print
24382         not interesting value histograms.
24384 2018-01-23  Martin Liska  <mliska@suse.cz>
24386         * profile-count.h (enum profile_quality): Add
24387         profile_uninitialized as the first value. Do not number values
24388         as they are zero based.
24389         (profile_count::verify): Update sanity check.
24390         (profile_probability::verify): Likewise.
24392 2018-01-23  Nathan Sidwell  <nathan@acm.org>
24394         * doc/invoke.texi (ffor-scope): Deprecate.
24396 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
24398         PR tree-optimization/83510
24399         * domwalk.c (set_all_edges_as_executable): New function.
24400         (dom_walker::dom_walker): Convert bool param
24401         "skip_unreachable_blocks" to enum reachability.  Move setup of
24402         edge flags to set_all_edges_as_executable and only do it when
24403         reachability is REACHABLE_BLOCKS.
24404         * domwalk.h (enum dom_walker::reachability): New enum.
24405         (dom_walker::dom_walker): Convert bool param
24406         "skip_unreachable_blocks" to enum reachability.
24407         (set_all_edges_as_executable): New decl.
24408         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
24409         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
24410         "reachability".
24411         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
24412         but converting true to REACHABLE_BLOCKS.
24413         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
24414         * tree-vrp.c
24415         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
24416         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
24417         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
24418         REACHABLE_BLOCKS.
24419         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
24420         if check_all_array_refs will be called.
24422 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
24424         * tree.c (selftest::test_location_wrappers): Add more test
24425         coverage.
24427 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
24429         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
24430         (selftest::test_bit_in_range): Likewise.
24432 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
24434         PR testsuite/83888
24435         * doc/sourcebuild.texi (vect_float): Say that the selector
24436         only describes the situation when -funsafe-math-optimizations is on.
24437         (vect_float_strict): Document.
24439 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
24441         PR tree-optimization/83965
24442         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
24443         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
24444         instead of checking only for a reduction.
24445         (vect_recog_widen_sum_pattern): Likewise.
24447 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
24449         * predict.c (probably_never_executed): Only use precise profile info.
24450         (compute_function_frequency): Skip after inlining hack since we now
24451         have quality checking.
24453 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
24455         * profile-count.h (profile_probability::very_unlikely,
24456         profile_probability::unlikely, profile_probability::even): Set
24457         precision to guessed.
24459 2018-01-23  Richard Biener  <rguenther@suse.de>
24461         PR tree-optimization/83963
24462         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
24463         Properly terminate dominator walk when crossing the exit edge not
24464         when visiting its source block.
24466 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
24468         PR c++/83918
24469         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
24470         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
24472 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
24474         PR tree-optimization/83957
24475         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
24476         semicolon after for body surrounded by braces.
24478         PR tree-optimization/83081
24479         * profile-count.h (profile_probability::split): New method.
24480         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
24481         Use profile_probability::split.
24482         (do_compare_rtx_and_jump): Fix adjustment of probabilities
24483         when splitting a single conditional jump into 2.
24485 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
24487         PR tree-optimization/69452
24488         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
24489         decl.
24491 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24493         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
24494         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
24495         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
24497 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24499         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
24500         declaration.
24501         * config/rl78/rl78.md (movdi): New define_expand.
24502         * config/rl78/rl78.c (rl78_split_movdi): New function.
24504 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
24506         PR target/83862
24507         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
24508         no longer used.
24509         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
24510         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
24511         128-bit to produce an UNSPEC move to get the double word with the
24512         signbit and then a shift directly to do signbit.
24513         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
24514         implementation with a new version that just does either a direct
24515         move or a regular move.  Move memory interface to separate insns.
24516         Move insns so they are next to the expander.
24517         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
24518         with signbit move.  Split big and little endian case.
24519         (signbit<mode>2_dm_mem_le): Likewise.
24520         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
24521         (signbit<mode>2_dm2): Likewise.
24523 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24525         * config/rl78/rl78.md (anddi3): New define_expand.
24527 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24529         * config/rl78/rl78.md (umindi3): New define_expand.
24531 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24533         * config/rl78/rl78.md (smindi3): New define_expand.
24535 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24537         * config/rl78/rl78.md (smaxdi3): New define_expand.
24539 2018-01-22  Carl Love  <cel@us.ibm.com>
24541         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
24542         LVX_V1TI): Add macro expansion.
24543         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
24544         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
24545         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
24546         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
24547         Change check to determine if the instruction is a byte reversing
24548         entry.  Fix typo in comment.
24549         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
24550         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
24551         Add def_builtin calls for new builtins.
24552         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
24553         Add define_insn expansion.
24555 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24557         * config/rl78/rl78.md (umaxdi3): New define_expand.
24559 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
24561         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
24562         for non-QImode registers.
24564 2018-01-22  Richard Biener  <rguenther@suse.de>
24566         PR tree-optimization/83963
24567         * graphite-scop-detection.c (scop_detection::get_sese): Delay
24568         including the loop exit block.
24569         (scop_detection::merge_sese): Likewise.
24570         (scop_detection::add_scop): Do it here instead.
24572 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24574         * doc/sourcebuild.texi (arm_softfloat): Document.
24576 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
24578         PR gcc/77734
24579         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
24580         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
24581         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
24583 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24584             David Edelsohn  <dje.gcc@gmail.com>
24586         PR target/83946
24587         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24588         Change "crset eq" to "crset 2".
24589         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24590         (*call_indirect_aix<mode>_nospec): Likewise.
24591         (*call_value_indirect_aix<mode>_nospec): Likewise.
24592         (*call_indirect_elfv2<mode>_nospec): Likewise.
24593         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
24594         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
24595         change assembly output from . to $.
24596         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24597         (indirect_jump<mode>_nospec): Change assembly output from . to $.
24598         (*tablejump<mode>_internal1_nospec): Likewise.
24600 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
24602         PR target/80870
24603         * config/sh/sh_optimize_sett_clrt.cc:
24604         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
24606 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
24608         PR tree-optimization/83940
24609         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
24610         offset_dt to vect_constant_def rather than vect_unknown_def_type.
24611         (vect_check_load_store_mask): Add a mask_dt_out parameter and
24612         use it to pass back the definition type.
24613         (vect_check_store_rhs): Likewise rhs_dt_out.
24614         (vect_build_gather_load_calls): Add a mask_dt argument and use
24615         it instead of a call to vect_is_simple_use.
24616         (vectorizable_store): Update calls to vect_check_load_store_mask
24617         and vect_check_store_rhs.  Use the dt returned by the latter instead
24618         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
24619         instead of calls to vect_is_simple_use.  Pass the scalar rather
24620         than the vector operand to vect_is_simple_use when handling
24621         second and subsequent copies of an rhs value.
24622         (vectorizable_load): Update calls to vect_check_load_store_mask
24623         and vect_build_gather_load_calls.  Use the cached mask_dt and
24624         gs_info.offset_dt instead of calls to vect_is_simple_use.
24626 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
24628         PR middle-end/83945
24629         * tree-emutls.c: Include gimplify.h.
24630         (lower_emutls_2): New function.
24631         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
24632         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
24633         it before further processing.
24635         PR target/83930
24636         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
24637         UINTVAL (trueop1) instead of INTVAL (op1).
24639 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
24641         PR debug/81570
24642         PR debug/83728
24643         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
24644         INCOMING_FRAME_SP_OFFSET if not defined.
24645         (scan_trace): Add ENTRY argument.  If true and
24646         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
24647         emit a note to adjust the CFA offset.
24648         (create_cfi_notes): Adjust scan_trace callers.
24649         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
24650         INCOMING_FRAME_SP_OFFSET in the CIE.
24651         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
24652         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
24653         Likewise.
24654         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
24655         * doc/tm.texi: Regenerated.
24657 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24659         PR rtl-optimization/83147
24660         * lra-constraints.c (remove_inheritance_pseudos): Use
24661         lra_substitute_pseudo_within_insn.
24663 2018-01-19  Tom de Vries  <tom@codesourcery.com>
24664             Cesar Philippidis  <cesar@codesourcery.com>
24666         PR target/83920
24667         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
24669 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
24671         PR target/83790
24672         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
24673         spaces for function labels.
24675 2018-01-19  Martin Liska  <mliska@suse.cz>
24677         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
24678         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
24679         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
24680         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
24681         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
24682         (PRED_CONST_RETURN): Change from 69 to 65.
24683         (PRED_NULL_RETURN): Change from 91 to 71.
24684         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
24685         (PRED_LOOP_GUARD): Change from 66 to 73.
24687 2018-01-19  Martin Liska  <mliska@suse.cz>
24689         * predict.c (predict_insn_def): Add new assert.
24690         (struct branch_predictor): Change type to signed integer.
24691         (test_prediction_value_range): Amend test to cover
24692         PROB_UNINITIALIZED.
24693         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
24694         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
24695         (PRED_LOOP_ITERATIONS_MAX): Likewise.
24696         (PRED_LOOP_IV_COMPARE): Likewise.
24697         * predict.h (PROB_UNINITIALIZED): Define new constant.
24699 2018-01-19  Martin Liska  <mliska@suse.cz>
24701         * predict.c (dump_prediction): Add new format for
24702         analyze_brprob.py script which is enabled with -details
24703         suboption.
24704         * profile-count.h (precise_p): New function.
24706 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
24708         PR tree-optimization/83922
24709         * tree-vect-loop.c (vect_verify_full_masking): Return false if
24710         there are no statements that need masking.
24711         (vect_active_double_reduction_p): New function.
24712         (vect_analyze_loop_operations): Use it when handling phis that
24713         are not in the loop header.
24715 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
24717         PR tree-optimization/83914
24718         * tree-vect-loop.c (vectorizable_induction): Don't convert
24719         init_expr or apply the peeling adjustment for inductions
24720         that are nested within the vectorized loop.
24722 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24724         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
24725         instead of NEG.
24727 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
24729         PR sanitizer/81715
24730         PR testsuite/83882
24731         * function.h (gimplify_parameters): Add gimple_seq * argument.
24732         * function.c: Include gimple.h and options.h.
24733         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
24734         for the added local temporaries if needed.
24735         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
24736         if there are any parameter cleanups, wrap whole body into a
24737         try/finally with the cleanups.
24739 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
24741         PR target/82964
24742         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
24743         Use GET_MODE_CLASS for scalar floating point.
24745 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
24747         PR ipa/82256
24748         patch by PaX Team
24749         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
24750         Fix call of call_cgraph_insertion_hooks.
24752 2018-01-18  Martin Sebor  <msebor@redhat.com>
24754         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
24756 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
24758         PR ipa/83619
24759         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
24760         frequencies.
24762 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
24764         PR other/70268
24765         * common.opt: (-ffile-prefix-map): New option.
24766         * opts.c (common_handle_option): Defer it.
24767         * opts-global.c (handle_common_deferred_options): Handle it.
24768         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
24769         * file-prefix-map.h: New file.
24770         (remap_debug_filename, add_debug_prefix_map): ...here.
24771         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
24772         * final.c (debug_prefix_map, add_debug_prefix_map
24773         remap_debug_filename): Move to...
24774         * file-prefix-map.c: New file.
24775         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
24776         generalize, get rid of alloca(), use strrchr() instead of strchr().
24777         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
24778         Implement in terms of add_prefix_map().
24779         (remap_macro_filename, remap_debug_filename): Implement in term of
24780         remap_filename().
24781         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
24782         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
24783         * dbxout.c: Include file-prefix-map.h.
24784         * varasm.c: Likewise.
24785         * vmsdbgout.c: Likewise.
24786         * xcoffout.c: Likewise.
24787         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
24788         * doc/cppopts.texi (-fmacro-prefix-map): Document.
24789         * doc/invoke.texi (-ffile-prefix-map): Document.
24790         (-fdebug-prefix-map): Update description.
24792 2018-01-18  Martin Liska  <mliska@suse.cz>
24794         * config/i386/i386.c (indirect_thunk_name): Document that also
24795         lfence is emitted.
24796         (output_indirect_thunk): Document why both instructions
24797         (pause and lfence) are generated.
24799 2018-01-18  Richard Biener  <rguenther@suse.de>
24801         PR tree-optimization/83887
24802         * graphite-scop-detection.c
24803         (scop_detection::get_nearest_dom_with_single_entry): Remove.
24804         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
24805         (scop_detection::merge_sese): Re-implement with a flood-fill
24806         algorithm that properly finds a SESE region if it exists.
24808 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
24810         PR c/61240
24811         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
24812         pointer_diff optimizations use view_convert instead of convert.
24814 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24816         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
24817         Generate different code for -mno-speculate-indirect-jumps.
24818         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
24819         (*call_indirect_aix<mode>): Disable for
24820         -mno-speculate-indirect-jumps.
24821         (*call_indirect_aix<mode>_nospec): New define_insn.
24822         (*call_value_indirect_aix<mode>): Disable for
24823         -mno-speculate-indirect-jumps.
24824         (*call_value_indirect_aix<mode>_nospec): New define_insn.
24825         (*sibcall_nonlocal_sysv<mode>): Generate different code for
24826         -mno-speculate-indirect-jumps.
24827         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
24829 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
24831         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
24832         long double type, set the flags for noting the default long double
24833         type, even if we don't pass or return a long double type.
24835 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
24837         PR ipa/83051
24838         * ipa-inline.c (flatten_function): Do not overwrite final inlining
24839         failure.
24841 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
24843         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
24844         support for merge[hl].
24845         (fold_mergehl_helper): New helper function.
24846         (tree-vector-builder.h): New #include for tree_vector_builder usage.
24847         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
24848         (altivec_vmrglw_direct): Add xxmrglw insn.
24850 2018-01-17  Andrew Waterman  <andrew@sifive.com>
24852         * config/riscv/riscv.c (riscv_conditional_register_usage): If
24853         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
24855 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
24857         PR lto/83121
24858         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
24859         call the lto_location_cache before reading the
24860         DECL_SOURCE_LOCATION of the types.
24862 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
24863             Richard Sandiford  <richard.sandiford@linaro.org>
24865         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
24866         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
24867         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
24868         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
24869         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
24870         Add declaration.
24871         * config/aarch64/constraints.md (aarch64_movti_operand):
24872         Limit immediates.
24873         * config/aarch64/predicates.md (Uti): Add new constraint.
24875 2018-01-17  Carl Love  <cel@us.ibm.com>
24877         * config/rs6000/vsx.md (define_expand xl_len_r,
24878         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
24879         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
24880         lxvll.
24881         (define_expand, define_insn): Move the shift left from  the
24882         define_insn to the define_expand for lxvl and stxvl instructions.
24883         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
24884         and XL_LEN_R definitions to PURE.
24886 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
24888         * config/i386/i386.c (indirect_thunk_name): Declare regno
24889         as unsigned int.  Compare regno with INVALID_REGNUM.
24890         (output_indirect_thunk): Ditto.
24891         (output_indirect_thunk_function): Ditto.
24892         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
24893         in the call to output_indirect_thunk_function.
24895 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
24897         PR middle-end/83884
24898         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
24899         rather than the size of inner_type to determine the stack slot size
24900         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
24902 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
24904         PR target/83546
24905         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
24906         to PTA_SILVERMONT.
24908 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
24910         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
24911         endian Linux systems to optionally enable multilibs for selecting
24912         the long double type if the user configured an explicit type.
24913         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
24914         have no long double multilibs if not defined.
24915         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
24916         warn if the user used -mabi={ieee,ibm}longdouble and we built
24917         multilibs for long double.
24918         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
24919         appropriate multilib option.
24920         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
24921         multilib options.
24922         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
24923         for building long double multilibs.
24924         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
24926 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
24928         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
24929         copies.
24931         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
24932         64 bits.
24933         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
24934         128 bits.
24936         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
24937         variables.
24939         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
24940         return value.
24942 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
24944         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
24945         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
24947 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24949         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
24950         different rtl trees depending on TARGET_64BIT.
24951         (rs6000_gen_lvx): Likewise.
24953 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
24955         * config/visium/visium.md (nop): Tweak comment.
24956         (hazard_nop): Likewise.
24958 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24960         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
24961         -mspeculate-indirect-jumps.
24962         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
24963         for -mno-speculate-indirect-jumps.
24964         (*call_indirect_elfv2<mode>_nospec): New define_insn.
24965         (*call_value_indirect_elfv2<mode>): Disable for
24966         -mno-speculate-indirect-jumps.
24967         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
24968         (indirect_jump): Emit different RTL for
24969         -mno-speculate-indirect-jumps.
24970         (*indirect_jump<mode>): Disable for
24971         -mno-speculate-indirect-jumps.
24972         (*indirect_jump<mode>_nospec): New define_insn.
24973         (tablejump): Emit different RTL for
24974         -mno-speculate-indirect-jumps.
24975         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
24976         (tablejumpsi_nospec): New define_expand.
24977         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
24978         (tablejumpdi_nospec): New define_expand.
24979         (*tablejump<mode>_internal1): Disable for
24980         -mno-speculate-indirect-jumps.
24981         (*tablejump<mode>_internal1_nospec): New define_insn.
24982         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
24983         option.
24985 2018-01-16  Artyom Skrobov tyomitch@gmail.com
24987         * caller-save.c (insert_save): Drop unnecessary parameter.  All
24988         callers updated.
24990 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
24991             Richard Biener  <rguenth@suse.de>
24993         PR libgomp/83590
24994         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
24995         return early, inline manually is_gimple_sizepos.  Make sure if we
24996         call gimplify_expr we don't end up with a gimple constant.
24997         * tree.c (variably_modified_type_p): Don't return true for
24998         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
24999         * gimplify.h (is_gimple_sizepos): Remove.
25001 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25003         PR tree-optimization/83857
25004         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
25005         vectorizable_live_operation for pure SLP statements.
25006         (vectorizable_live_operation): Handle PHIs.
25008 2018-01-16  Richard Biener  <rguenther@suse.de>
25010         PR tree-optimization/83867
25011         * tree-vect-stmts.c (vect_transform_stmt): Precompute
25012         nested_in_vect_loop_p since the scalar stmt may get invalidated.
25014 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25016         PR c/83844
25017         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
25018         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
25019         If off is not INTEGER_CST, issue a may not be aligned warning
25020         rather than isn't aligned.  Use isn%'t rather than isn't.
25021         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
25022         into MULT_EXPR.
25023         <case MULT_EXPR>: Improve the case when bottom and one of the
25024         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
25025         operand, in that case check if the other operand is multiple of
25026         bottom divided by the INTEGER_CST operand.
25028 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25030         PR target/83858
25031         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
25032         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
25033         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
25034         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
25035         * config/pa/pa.c (pa_function_arg_advance): Likewise.
25036         (pa_function_arg, pa_arg_partial_bytes): Likewise.
25037         (pa_function_arg_size): New function.
25039 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25041         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
25042         in a separate statement.
25044 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
25046         PR tree-optimization/83847
25047         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
25048         group gathers and scatters.
25050 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
25052         PR rtl-optimization/86620
25053         * params.def (max-sched-ready-insns): Bump minimum value to 1.
25055         PR rtl-optimization/83213
25056         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
25057         to last if both are JUMP_INSNs.
25059         PR tree-optimization/83843
25060         * gimple-ssa-store-merging.c
25061         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
25062         store_immediate_info for bswap/nop orig_stores.
25064 2018-01-15  Andrew Waterman  <andrew@sifive.com>
25066         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
25067         !TARGET_MUL.
25068         <UDIV>: Increase cost if !TARGET_DIV.
25070 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
25072         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
25073         (define_attr "cr_logical_3op"): New.
25074         (cceq_ior_compare): Adjust.
25075         (cceq_ior_compare_complement): Adjust.
25076         (*cceq_rev_compare): Adjust.
25077         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
25078         (is_cracked_insn): Adjust.
25079         (insn_must_be_first_in_group): Adjust.
25080         * config/rs6000/40x.md: Adjust.
25081         * config/rs6000/440.md: Adjust.
25082         * config/rs6000/476.md: Adjust.
25083         * config/rs6000/601.md: Adjust.
25084         * config/rs6000/603.md: Adjust.
25085         * config/rs6000/6xx.md: Adjust.
25086         * config/rs6000/7450.md: Adjust.
25087         * config/rs6000/7xx.md: Adjust.
25088         * config/rs6000/8540.md: Adjust.
25089         * config/rs6000/cell.md: Adjust.
25090         * config/rs6000/e300c2c3.md: Adjust.
25091         * config/rs6000/e500mc.md: Adjust.
25092         * config/rs6000/e500mc64.md: Adjust.
25093         * config/rs6000/e5500.md: Adjust.
25094         * config/rs6000/e6500.md: Adjust.
25095         * config/rs6000/mpc.md: Adjust.
25096         * config/rs6000/power4.md: Adjust.
25097         * config/rs6000/power5.md: Adjust.
25098         * config/rs6000/power6.md: Adjust.
25099         * config/rs6000/power7.md: Adjust.
25100         * config/rs6000/power8.md: Adjust.
25101         * config/rs6000/power9.md: Adjust.
25102         * config/rs6000/rs64.md: Adjust.
25103         * config/rs6000/titan.md: Adjust.
25105 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25107         * config/i386/predicates.md (indirect_branch_operand): Rewrite
25108         ix86_indirect_branch_register logic.
25110 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25112         * config/i386/constraints.md (Bs): Update
25113         ix86_indirect_branch_register check.  Don't check
25114         ix86_indirect_branch_register with GOT_memory_operand.
25115         (Bw): Likewise.
25116         * config/i386/predicates.md (GOT_memory_operand): Don't check
25117         ix86_indirect_branch_register here.
25118         (GOT32_symbol_operand): Likewise.
25120 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25122         * config/i386/predicates.md (constant_call_address_operand):
25123         Rewrite ix86_indirect_branch_register logic.
25124         (sibcall_insn_operand): Likewise.
25126 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25128         * config/i386/constraints.md (Bs): Replace
25129         ix86_indirect_branch_thunk_register with
25130         ix86_indirect_branch_register.
25131         (Bw): Likewise.
25132         * config/i386/i386.md (indirect_jump): Likewise.
25133         (tablejump): Likewise.
25134         (*sibcall_memory): Likewise.
25135         (*sibcall_value_memory): Likewise.
25136         Peepholes of indirect call and jump via memory: Likewise.
25137         * config/i386/i386.opt: Likewise.
25138         * config/i386/predicates.md (indirect_branch_operand): Likewise.
25139         (GOT_memory_operand): Likewise.
25140         (call_insn_operand): Likewise.
25141         (sibcall_insn_operand): Likewise.
25142         (GOT32_symbol_operand): Likewise.
25144 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
25146         PR middle-end/83837
25147         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
25148         type rather than type addr's type points to.
25149         (expand_omp_atomic_mutex): Likewise.
25150         (expand_omp_atomic): Likewise.
25152 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
25154         PR target/83839
25155         * config/i386/i386.c (output_indirect_thunk_function): Use
25156         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
25157         for  __x86_return_thunk.
25159 2018-01-15  Richard Biener  <rguenther@suse.de>
25161         PR middle-end/83850
25162         * expmed.c (extract_bit_field_1): Fix typo.
25164 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25166         PR target/83687
25167         * config/arm/iterators.md (VF): New mode iterator.
25168         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
25169         Remove integer-related logic from pattern.
25170         (neon_vabd<mode>_3): Likewise.
25172 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
25174         PR middle-end/82694
25175         * common.opt (fstrict-overflow): No longer an alias.
25176         (fwrapv-pointer): New option.
25177         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
25178         also for pointer types based on flag_wrapv_pointer.
25179         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
25180         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
25181         opts->x_flag_wrapv got set.
25182         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
25183         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
25184         POINTER_TYPE_OVERFLOW_UNDEFINED.
25185         * match.pd: Likewise in address comparison pattern.
25186         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
25188 2018-01-15  Richard Biener  <rguenther@suse.de>
25190         PR lto/83804
25191         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
25192         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
25193         Reset type names to their identifier if their TYPE_DECL doesn't
25194         have linkage (and thus is used for ODR and devirt).
25195         (save_debug_info_for_decl): Remove.
25196         (save_debug_info_for_type): Likewise.
25197         (add_tree_to_fld_list): Adjust.
25198         * tree-pretty-print.c (dump_generic_node): Make dumping of
25199         type names more robust.
25201 2018-01-15  Richard Biener  <rguenther@suse.de>
25203         * BASE-VER: Bump to 8.0.1.
25205 2018-01-14  Martin Sebor  <msebor@redhat.com>
25207         PR other/83508
25208         * builtins.c (check_access): Avoid warning when the no-warning bit
25209         is set.
25211 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
25213         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
25214         * ira-color (allocno_hard_regs_compare): Likewise.
25216 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
25218         PR target/83013
25219         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
25220         Use .pushsection/.popsection.
25222 2018-01-14  Martin Sebor  <msebor@redhat.com>
25224         PR c++/81327
25225         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
25227 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
25229         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
25230         entry from extra_headers.
25231         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
25232         extra_headers, make the list bitwise identical to the i?86-*-* one.
25234 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25236         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
25237         -mcmodel=large with -mindirect-branch=thunk,
25238         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
25239         -mfunction-return=thunk-extern.
25240         * doc/invoke.texi: Document -mcmodel=large is incompatible with
25241         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
25242         -mfunction-return=thunk and -mfunction-return=thunk-extern.
25244 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25246         * config/i386/i386.c (print_reg): Print the name of the full
25247         integer register without '%'.
25248         (ix86_print_operand): Handle 'V'.
25249         * doc/extend.texi: Document 'V' modifier.
25251 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25253         * config/i386/constraints.md (Bs): Disallow memory operand for
25254         -mindirect-branch-register.
25255         (Bw): Likewise.
25256         * config/i386/predicates.md (indirect_branch_operand): Likewise.
25257         (GOT_memory_operand): Likewise.
25258         (call_insn_operand): Likewise.
25259         (sibcall_insn_operand): Likewise.
25260         (GOT32_symbol_operand): Likewise.
25261         * config/i386/i386.md (indirect_jump): Call convert_memory_address
25262         for -mindirect-branch-register.
25263         (tablejump): Likewise.
25264         (*sibcall_memory): Likewise.
25265         (*sibcall_value_memory): Likewise.
25266         Disallow peepholes of indirect call and jump via memory for
25267         -mindirect-branch-register.
25268         (*call_pop): Replace m with Bw.
25269         (*call_value_pop): Likewise.
25270         (*sibcall_pop_memory): Replace m with Bs.
25271         * config/i386/i386.opt (mindirect-branch-register): New option.
25272         * doc/invoke.texi: Document -mindirect-branch-register option.
25274 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25276         * config/i386/i386-protos.h (ix86_output_function_return): New.
25277         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
25278         set function_return_type.
25279         (indirect_thunk_name): Add ret_p to indicate thunk for function
25280         return.
25281         (output_indirect_thunk_function): Pass false to
25282         indirect_thunk_name.
25283         (ix86_output_indirect_branch_via_reg): Likewise.
25284         (ix86_output_indirect_branch_via_push): Likewise.
25285         (output_indirect_thunk_function): Create alias for function
25286         return thunk if regno < 0.
25287         (ix86_output_function_return): New function.
25288         (ix86_handle_fndecl_attribute): Handle function_return.
25289         (ix86_attribute_table): Add function_return.
25290         * config/i386/i386.h (machine_function): Add
25291         function_return_type.
25292         * config/i386/i386.md (simple_return_internal): Use
25293         ix86_output_function_return.
25294         (simple_return_internal_long): Likewise.
25295         * config/i386/i386.opt (mfunction-return=): New option.
25296         (indirect_branch): Mention -mfunction-return=.
25297         * doc/extend.texi: Document function_return function attribute.
25298         * doc/invoke.texi: Document -mfunction-return= option.
25300 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
25302         * config/i386/i386-opts.h (indirect_branch): New.
25303         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
25304         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
25305         with local indirect jump when converting indirect call and jump.
25306         (ix86_set_indirect_branch_type): New.
25307         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
25308         (indirectlabelno): New.
25309         (indirect_thunk_needed): Likewise.
25310         (indirect_thunk_bnd_needed): Likewise.
25311         (indirect_thunks_used): Likewise.
25312         (indirect_thunks_bnd_used): Likewise.
25313         (INDIRECT_LABEL): Likewise.
25314         (indirect_thunk_name): Likewise.
25315         (output_indirect_thunk): Likewise.
25316         (output_indirect_thunk_function): Likewise.
25317         (ix86_output_indirect_branch_via_reg): Likewise.
25318         (ix86_output_indirect_branch_via_push): Likewise.
25319         (ix86_output_indirect_branch): Likewise.
25320         (ix86_output_indirect_jmp): Likewise.
25321         (ix86_code_end): Call output_indirect_thunk_function if needed.
25322         (ix86_output_call_insn): Call ix86_output_indirect_branch if
25323         needed.
25324         (ix86_handle_fndecl_attribute): Handle indirect_branch.
25325         (ix86_attribute_table): Add indirect_branch.
25326         * config/i386/i386.h (machine_function): Add indirect_branch_type
25327         and has_local_indirect_jump.
25328         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
25329         to true.
25330         (tablejump): Likewise.
25331         (*indirect_jump): Use ix86_output_indirect_jmp.
25332         (*tablejump_1): Likewise.
25333         (simple_return_indirect_internal): Likewise.
25334         * config/i386/i386.opt (mindirect-branch=): New option.
25335         (indirect_branch): New.
25336         (keep): Likewise.
25337         (thunk): Likewise.
25338         (thunk-inline): Likewise.
25339         (thunk-extern): Likewise.
25340         * doc/extend.texi: Document indirect_branch function attribute.
25341         * doc/invoke.texi: Document -mindirect-branch= option.
25343 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
25345         PR ipa/83051
25346         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
25348 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
25350         * ipa-inline.c (want_inline_small_function_p): Return false if
25351         inlining has already failed with CIF_FINAL_ERROR.
25352         (update_caller_keys): Call want_inline_small_function_p before
25353         can_inline_edge_p.
25354         (update_callee_keys): Likewise.
25356 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25358         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
25359         New function.
25360         (rs6000_quadword_masked_address_p): Likewise.
25361         (quad_aligned_load_p): Likewise.
25362         (quad_aligned_store_p): Likewise.
25363         (const_load_sequence_p): Add comment to describe the outer-most loop.
25364         (mimic_memory_attributes_and_flags): New function.
25365         (rs6000_gen_stvx): Likewise.
25366         (replace_swapped_aligned_store): Likewise.
25367         (rs6000_gen_lvx): Likewise.
25368         (replace_swapped_aligned_load): Likewise.
25369         (replace_swapped_load_constant): Capitalize argument name in
25370         comment describing this function.
25371         (rs6000_analyze_swaps): Add a third pass to search for vector loads
25372         and stores that access quad-word aligned addresses and replace
25373         with stvx or lvx instructions when appropriate.
25374         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
25375         New function prototype.
25376         (rs6000_quadword_masked_address_p): Likewise.
25377         (rs6000_gen_lvx): Likewise.
25378         (rs6000_gen_stvx): Likewise.
25379         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
25380         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
25381         when memory address is aligned.
25382         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
25383         this split to select lvx instruction when memory address is aligned.
25384         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
25385         instruction when memory address is aligned.
25386         (*vsx_le_perm_load_v16qi): Likewise.
25387         (four unnamed splitters): Modify to select the stvx instruction
25388         when memory is aligned.
25390 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
25392         * predict.c (determine_unlikely_bbs): Handle correctly BBs
25393         which appears in the queue multiple times.
25395 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25396             Alan Hayward  <alan.hayward@arm.com>
25397             David Sherwood  <david.sherwood@arm.com>
25399         * tree-vectorizer.h (vec_lower_bound): New structure.
25400         (_loop_vec_info): Add check_nonzero and lower_bounds.
25401         (LOOP_VINFO_CHECK_NONZERO): New macro.
25402         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
25403         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
25404         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
25405         fields.  Make seg_len the distance travelled, not including the
25406         access size.
25407         (dr_direction_indicator): Declare.
25408         (dr_zero_step_indicator): Likewise.
25409         (dr_known_forward_stride_p): Likewise.
25410         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
25411         tree-ssanames.h.
25412         (runtime_alias_check_p): Allow runtime alias checks with
25413         variable strides.
25414         (operator ==): Compare access_size and align.
25415         (prune_runtime_alias_test_list): Rework for new distinction between
25416         the access_size and seg_len.
25417         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
25418         segment lengths.
25419         (get_segment_min_max): New function.
25420         (create_intersect_range_checks): Use it.
25421         (dr_step_indicator): New function.
25422         (dr_direction_indicator): Likewise.
25423         (dr_zero_step_indicator): Likewise.
25424         (dr_known_forward_stride_p): Likewise.
25425         * tree-loop-distribution.c (data_ref_segment_size): Return
25426         DR_STEP * (niters - 1).
25427         (compute_alias_check_pairs): Update call to the dr_with_seg_len
25428         constructor.
25429         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
25430         (vect_preserves_scalar_order_p): New function, split out from...
25431         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
25432         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
25433         (vect_vfa_access_size): New function.
25434         (vect_vfa_align): Likewise.
25435         (vect_compile_time_alias): Take access_size_a and access_b arguments.
25436         (dump_lower_bound): New function.
25437         (vect_check_lower_bound): Likewise.
25438         (vect_small_gap_p): Likewise.
25439         (vectorizable_with_step_bound_p): Likewise.
25440         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
25441         depencies if the vectorization factor is 1.  Convert the checks
25442         for nonzero steps into checks on the bounds of DR_STEP.  Try using
25443         a bunds check for variable steps if the minimum required step is
25444         relatively small. Update calls to the dr_with_seg_len
25445         constructor and to vect_compile_time_alias.
25446         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
25447         function.
25448         (vect_loop_versioning): Call it.
25449         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
25450         when retrying.
25451         (vect_estimate_min_profitable_iters): Account for any bounds checks.
25453 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25454             Alan Hayward  <alan.hayward@arm.com>
25455             David Sherwood  <david.sherwood@arm.com>
25457         * doc/sourcebuild.texi (vect_scatter_store): Document.
25458         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
25459         optabs.
25460         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
25461         Document.
25462         * genopinit.c (main): Add supports_vec_scatter_store and
25463         supports_vec_scatter_store_cached to target_optabs.
25464         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
25465         IFN_MASK_SCATTER_STORE.
25466         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
25467         functions.
25468         * internal-fn.h (internal_store_fn_p): Declare.
25469         (internal_fn_stored_value_index): Likewise.
25470         * internal-fn.c (scatter_store_direct): New macro.
25471         (expand_scatter_store_optab_fn): New function.
25472         (direct_scatter_store_optab_supported_p): New macro.
25473         (internal_store_fn_p): New function.
25474         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
25475         IFN_MASK_SCATTER_STORE.
25476         (internal_fn_mask_index): Likewise.
25477         (internal_fn_stored_value_index): New function.
25478         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
25479         for scatter stores.
25480         * optabs-query.h (supports_vec_scatter_store_p): Declare.
25481         * optabs-query.c (supports_vec_scatter_store_p): New function.
25482         * tree-vectorizer.h (vect_get_store_rhs): Declare.
25483         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
25484         true for scatter stores.
25485         (vect_gather_scatter_fn_p): Handle scatter stores too.
25486         (vect_check_gather_scatter): Consider using scatter stores if
25487         supports_vec_scatter_store_p.
25488         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
25489         scatter stores too.
25490         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
25491         internal_fn_stored_value_index.
25492         (check_load_store_masking): Handle scatter stores too.
25493         (vect_get_store_rhs): Make public.
25494         (vectorizable_call): Use internal_store_fn_p.
25495         (vectorizable_store): Handle scatter store internal functions.
25496         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
25497         when deciding whether the end of the group has been reached.
25498         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
25499         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
25500         (mask_scatter_store<mode>): New insns.
25502 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25503             Alan Hayward  <alan.hayward@arm.com>
25504             David Sherwood  <david.sherwood@arm.com>
25506         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
25507         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
25508         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
25509         function.
25510         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
25511         Use vect_truncate_gather_scatter_offset if we can't treat the
25512         operation as a normal gather load or scatter store.
25513         (get_group_load_store_type): Take the gather_scatter_info
25514         as argument.  Try using a gather load or scatter store for
25515         single-element groups.
25516         (get_load_store_type): Update calls to get_group_load_store_type
25517         and vect_use_strided_gather_scatters_p.
25519 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25520             Alan Hayward  <alan.hayward@arm.com>
25521             David Sherwood  <david.sherwood@arm.com>
25523         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
25524         optional tree argument.
25525         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
25526         null target hooks.
25527         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
25528         but continue to use the current value as a fallback.
25529         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
25530         to compare the updates.
25531         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
25532         (get_load_store_type): Use it when handling a strided access.
25533         (vect_get_strided_load_store_ops): New function.
25534         (vect_get_data_ptr_increment): Likewise.
25535         (vectorizable_load): Handle strided gather loads.  Always pass
25536         a step to vect_create_data_ref_ptr and bump_vector_ptr.
25538 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25539             Alan Hayward  <alan.hayward@arm.com>
25540             David Sherwood  <david.sherwood@arm.com>
25542         * doc/md.texi (gather_load@var{m}): Document.
25543         (mask_gather_load@var{m}): Likewise.
25544         * genopinit.c (main): Add supports_vec_gather_load and
25545         supports_vec_gather_load_cached to target_optabs.
25546         * optabs-tree.c (init_tree_optimization_optabs): Use
25547         ggc_cleared_alloc to allocate target_optabs.
25548         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
25549         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
25550         functions.
25551         * internal-fn.h (internal_load_fn_p): Declare.
25552         (internal_gather_scatter_fn_p): Likewise.
25553         (internal_fn_mask_index): Likewise.
25554         (internal_gather_scatter_fn_supported_p): Likewise.
25555         * internal-fn.c (gather_load_direct): New macro.
25556         (expand_gather_load_optab_fn): New function.
25557         (direct_gather_load_optab_supported_p): New macro.
25558         (direct_internal_fn_optab): New function.
25559         (internal_load_fn_p): Likewise.
25560         (internal_gather_scatter_fn_p): Likewise.
25561         (internal_fn_mask_index): Likewise.
25562         (internal_gather_scatter_fn_supported_p): Likewise.
25563         * optabs-query.c (supports_at_least_one_mode_p): New function.
25564         (supports_vec_gather_load_p): Likewise.
25565         * optabs-query.h (supports_vec_gather_load_p): Declare.
25566         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
25567         and memory_type field.
25568         (NUM_PATTERNS): Bump to 15.
25569         * tree-vect-data-refs.c: Include internal-fn.h.
25570         (vect_gather_scatter_fn_p): New function.
25571         (vect_describe_gather_scatter_call): Likewise.
25572         (vect_check_gather_scatter): Try using internal functions for
25573         gather loads.  Recognize existing calls to a gather load function.
25574         (vect_analyze_data_refs): Consider using gather loads if
25575         supports_vec_gather_load_p.
25576         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
25577         (vect_get_gather_scatter_offset_type): Likewise.
25578         (vect_convert_mask_for_vectype): Likewise.
25579         (vect_add_conversion_to_patterm): Likewise.
25580         (vect_try_gather_scatter_pattern): Likewise.
25581         (vect_recog_gather_scatter_pattern): New pattern recognizer.
25582         (vect_vect_recog_func_ptrs): Add it.
25583         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
25584         internal_fn_mask_index and internal_gather_scatter_fn_p.
25585         (check_load_store_masking): Take the gather_scatter_info as an
25586         argument and handle gather loads.
25587         (vect_get_gather_scatter_ops): New function.
25588         (vectorizable_call): Check internal_load_fn_p.
25589         (vectorizable_load): Likewise.  Handle gather load internal
25590         functions.
25591         (vectorizable_store): Update call to check_load_store_masking.
25592         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
25593         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
25594         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
25595         (aarch64_gather_scale_operand_d): New predicates.
25596         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
25597         (mask_gather_load<mode>): New insns.
25599 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25600             Alan Hayward  <alan.hayward@arm.com>
25601             David Sherwood  <david.sherwood@arm.com>
25603         * optabs.def (fold_left_plus_optab): New optab.
25604         * doc/md.texi (fold_left_plus_@var{m}): Document.
25605         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
25606         * internal-fn.c (fold_left_direct): Define.
25607         (expand_fold_left_optab_fn): Likewise.
25608         (direct_fold_left_optab_supported_p): Likewise.
25609         * fold-const-call.c (fold_const_fold_left): New function.
25610         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
25611         * tree-parloops.c (valid_reduction_p): New function.
25612         (gather_scalar_reductions): Use it.
25613         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
25614         (vect_finish_replace_stmt): Declare.
25615         * tree-vect-loop.c (fold_left_reduction_fn): New function.
25616         (needs_fold_left_reduction_p): New function, split out from...
25617         (vect_is_simple_reduction): ...here.  Accept reductions that
25618         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
25619         (vect_force_simple_reduction): Also store the reduction type in
25620         the assignment's STMT_VINFO_REDUC_TYPE.
25621         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
25622         (merge_with_identity): New function.
25623         (vect_expand_fold_left): Likewise.
25624         (vectorize_fold_left_reduction): Likewise.
25625         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
25626         scalar phi in place for it.  Check for target support and reject
25627         cases that would reassociate the operation.  Defer the transform
25628         phase to vectorize_fold_left_reduction.
25629         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
25630         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
25631         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
25633 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25635         * tree-if-conv.c (predicate_mem_writes): Remove redundant
25636         call to ifc_temp_var.
25638 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25639             Alan Hayward  <alan.hayward@arm.com>
25640             David Sherwood  <david.sherwood@arm.com>
25642         * target.def (legitimize_address_displacement): Take the original
25643         offset as a poly_int.
25644         * targhooks.h (default_legitimize_address_displacement): Update
25645         accordingly.
25646         * targhooks.c (default_legitimize_address_displacement): Likewise.
25647         * doc/tm.texi: Regenerate.
25648         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
25649         as an argument, moving assert of ad->disp == ad->disp_term to...
25650         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
25651         Try calling targetm.legitimize_address_displacement before expanding
25652         the address rather than afterwards, and adjust for the new interface.
25653         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
25654         Match the new hook interface.  Handle SVE addresses.
25655         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
25656         new hook interface.
25658 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25660         * Makefile.in (OBJS): Add early-remat.o.
25661         * target.def (select_early_remat_modes): New hook.
25662         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
25663         * doc/tm.texi: Regenerate.
25664         * targhooks.h (default_select_early_remat_modes): Declare.
25665         * targhooks.c (default_select_early_remat_modes): New function.
25666         * timevar.def (TV_EARLY_REMAT): New timevar.
25667         * passes.def (pass_early_remat): New pass.
25668         * tree-pass.h (make_pass_early_remat): Declare.
25669         * early-remat.c: New file.
25670         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
25671         function.
25672         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
25674 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25675             Alan Hayward  <alan.hayward@arm.com>
25676             David Sherwood  <david.sherwood@arm.com>
25678         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
25679         vfm1 with a bound_epilog parameter.
25680         (vect_do_peeling): Update calls accordingly, and move the prologue
25681         call earlier in the function.  Treat the base bound_epilog as 0 for
25682         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
25683         this base when peeling for gaps.
25684         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
25685         with fully-masked loops.
25686         (vect_estimate_min_profitable_iters): Handle the single peeled
25687         iteration in that case.
25689 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25690             Alan Hayward  <alan.hayward@arm.com>
25691             David Sherwood  <david.sherwood@arm.com>
25693         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
25694         single-element interleaving even if the size is not a power of 2.
25695         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
25696         accesses for single-element interleaving if the group size is
25697         not a power of 2.
25699 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25700             Alan Hayward  <alan.hayward@arm.com>
25701             David Sherwood  <david.sherwood@arm.com>
25703         * doc/md.texi (fold_extract_last_@var{m}): Document.
25704         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
25705         * optabs.def (fold_extract_last_optab): New optab.
25706         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
25707         * internal-fn.c (fold_extract_direct): New macro.
25708         (expand_fold_extract_optab_fn): Likewise.
25709         (direct_fold_extract_optab_supported_p): Likewise.
25710         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
25711         * tree-vect-loop.c (vect_model_reduction_cost): Handle
25712         EXTRACT_LAST_REDUCTION.
25713         (get_initial_def_for_reduction): Do not create an initial vector
25714         for EXTRACT_LAST_REDUCTION reductions.
25715         (vectorizable_reduction): Leave the scalar phi in place for
25716         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
25717         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
25718         epilogue code for EXTRACT_LAST_REDUCTION and defer the
25719         transform phase to vectorizable_condition.
25720         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
25721         split out from...
25722         (vect_finish_stmt_generation): ...here.
25723         (vect_finish_replace_stmt): New function.
25724         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
25725         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
25726         pattern.
25727         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
25729 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25730             Alan Hayward  <alan.hayward@arm.com>
25731             David Sherwood  <david.sherwood@arm.com>
25733         * doc/md.texi (extract_last_@var{m}): Document.
25734         * optabs.def (extract_last_optab): New optab.
25735         * internal-fn.def (EXTRACT_LAST): New internal function.
25736         * internal-fn.c (cond_unary_direct): New macro.
25737         (expand_cond_unary_optab_fn): Likewise.
25738         (direct_cond_unary_optab_supported_p): Likewise.
25739         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
25740         loops using EXTRACT_LAST.
25741         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
25742         (extract_last_<mode>): ...this optab.
25743         (vec_extract<mode><Vel>): Update accordingly.
25745 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25746             Alan Hayward  <alan.hayward@arm.com>
25747             David Sherwood  <david.sherwood@arm.com>
25749         * target.def (empty_mask_is_expensive): New hook.
25750         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
25751         * doc/tm.texi: Regenerate.
25752         * targhooks.h (default_empty_mask_is_expensive): Declare.
25753         * targhooks.c (default_empty_mask_is_expensive): New function.
25754         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
25755         if the target says that empty masks are expensive.
25756         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
25757         New function.
25758         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
25760 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25761             Alan Hayward  <alan.hayward@arm.com>
25762             David Sherwood  <david.sherwood@arm.com>
25764         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
25765         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
25766         (vect_use_loop_mask_for_alignment_p): New function.
25767         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
25768         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
25769         niters_skip argument.  Make sure that the first niters_skip elements
25770         of the first iteration are inactive.
25771         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
25772         Update call to vect_set_loop_masks_directly.
25773         (get_misalign_in_elems): New function, split out from...
25774         (vect_gen_prolog_loop_niters): ...here.
25775         (vect_update_init_of_dr): Take a code argument that specifies whether
25776         the adjustment should be added or subtracted.
25777         (vect_update_init_of_drs): Likewise.
25778         (vect_prepare_for_masked_peels): New function.
25779         (vect_do_peeling): Skip prologue peeling if we're using a mask
25780         instead.  Update call to vect_update_inits_of_drs.
25781         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25782         mask_skip_niters.
25783         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
25784         alignment.  Do not include the number of peeled iterations in
25785         the minimum threshold in that case.
25786         (vectorizable_induction): Adjust the start value down by
25787         LOOP_VINFO_MASK_SKIP_NITERS iterations.
25788         (vect_transform_loop): Call vect_prepare_for_masked_peels.
25789         Take the number of skipped iterations into account when calculating
25790         the loop bounds.
25791         * tree-vect-stmts.c (vect_gen_while_not): New function.
25793 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25794             Alan Hayward  <alan.hayward@arm.com>
25795             David Sherwood  <david.sherwood@arm.com>
25797         * doc/sourcebuild.texi (vect_fully_masked): Document.
25798         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
25799         default value to 0.
25800         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
25801         split out from...
25802         (vect_analyze_loop_2): ...here. Don't check the vectorization
25803         factor against the number of loop iterations if the loop is
25804         fully-masked.
25806 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25807             Alan Hayward  <alan.hayward@arm.com>
25808             David Sherwood  <david.sherwood@arm.com>
25810         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
25811         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
25812         (dump_groups): Update accordingly.
25813         (iv_use::mem_type): New member variable.
25814         (address_p): New function.
25815         (record_use): Add a mem_type argument and initialize the new
25816         mem_type field.
25817         (record_group_use): Add a mem_type argument.  Use address_p.
25818         Remove obsolete null checks of base_object.  Update call to record_use.
25819         (find_interesting_uses_op): Update call to record_group_use.
25820         (find_interesting_uses_cond): Likewise.
25821         (find_interesting_uses_address): Likewise.
25822         (get_mem_type_for_internal_fn): New function.
25823         (find_address_like_use): Likewise.
25824         (find_interesting_uses_stmt): Try find_address_like_use before
25825         calling find_interesting_uses_op.
25826         (addr_offset_valid_p): Use the iv mem_type field as the type
25827         of the addressed memory.
25828         (add_autoinc_candidates): Likewise.
25829         (get_address_cost): Likewise.
25830         (split_small_address_groups_p): Use address_p.
25831         (split_address_groups): Likewise.
25832         (add_iv_candidate_for_use): Likewise.
25833         (autoinc_possible_for_pair): Likewise.
25834         (rewrite_groups): Likewise.
25835         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
25836         (determine_group_iv_cost): Update after split of USE_ADDRESS.
25837         (get_alias_ptr_type_for_ptr_address): New function.
25838         (rewrite_use_address): Rewrite address uses in calls that were
25839         identified by find_address_like_use.
25841 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25842             Alan Hayward  <alan.hayward@arm.com>
25843             David Sherwood  <david.sherwood@arm.com>
25845         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
25846         TARGET_MEM_REFs.
25847         * gimple-expr.h (is_gimple_addressable: Likewise.
25848         * gimple-expr.c (is_gimple_address): Likewise.
25849         * internal-fn.c (expand_call_mem_ref): New function.
25850         (expand_mask_load_optab_fn): Use it.
25851         (expand_mask_store_optab_fn): Likewise.
25853 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25854             Alan Hayward  <alan.hayward@arm.com>
25855             David Sherwood  <david.sherwood@arm.com>
25857         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
25858         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
25859         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
25860         (cond_umax@var{mode}): Document.
25861         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
25862         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
25863         (cond_umin_optab, cond_umax_optab): New optabs.
25864         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
25865         (COND_IOR, COND_XOR): New internal functions.
25866         * internal-fn.h (get_conditional_internal_fn): Declare.
25867         * internal-fn.c (cond_binary_direct): New macro.
25868         (expand_cond_binary_optab_fn): Likewise.
25869         (direct_cond_binary_optab_supported_p): Likewise.
25870         (get_conditional_internal_fn): New function.
25871         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
25872         Cope with reduction statements that are vectorized as calls rather
25873         than assignments.
25874         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
25875         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
25876         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
25877         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
25878         (UNSPEC_COND_EOR): New unspecs.
25879         (optab): Add mappings for them.
25880         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
25881         (sve_int_op, sve_fp_op): New int attributes.
25883 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25884             Alan Hayward  <alan.hayward@arm.com>
25885             David Sherwood  <david.sherwood@arm.com>
25887         * optabs.def (while_ult_optab): New optab.
25888         * doc/md.texi (while_ult@var{m}@var{n}): Document.
25889         * internal-fn.def (WHILE_ULT): New internal function.
25890         * internal-fn.h (direct_internal_fn_supported_p): New override
25891         that takes two types as argument.
25892         * internal-fn.c (while_direct): New macro.
25893         (expand_while_optab_fn): New function.
25894         (convert_optab_supported_p): Likewise.
25895         (direct_while_optab_supported_p): New macro.
25896         * wide-int.h (wi::udiv_ceil): New function.
25897         * tree-vectorizer.h (rgroup_masks): New structure.
25898         (vec_loop_masks): New typedef.
25899         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
25900         and fully_masked_p.
25901         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
25902         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
25903         (vect_max_vf): New function.
25904         (slpeel_make_loop_iterate_ntimes): Delete.
25905         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
25906         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
25907         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
25908         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
25909         internal-fn.h, stor-layout.h and optabs-query.h.
25910         (vect_set_loop_mask): New function.
25911         (add_preheader_seq): Likewise.
25912         (add_header_seq): Likewise.
25913         (interleave_supported_p): Likewise.
25914         (vect_maybe_permute_loop_masks): Likewise.
25915         (vect_set_loop_masks_directly): Likewise.
25916         (vect_set_loop_condition_masked): Likewise.
25917         (vect_set_loop_condition_unmasked): New function, split out from
25918         slpeel_make_loop_iterate_ntimes.
25919         (slpeel_make_loop_iterate_ntimes): Rename to..
25920         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
25921         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
25922         (vect_do_peeling): Update call accordingly.
25923         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
25924         loops.
25925         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25926         mask_compare_type, can_fully_mask_p and fully_masked_p.
25927         (release_vec_loop_masks): New function.
25928         (_loop_vec_info): Use it to free the loop masks.
25929         (can_produce_all_loop_masks_p): New function.
25930         (vect_get_max_nscalars_per_iter): Likewise.
25931         (vect_verify_full_masking): Likewise.
25932         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
25933         retries, and free the mask rgroups before retrying.  Check loop-wide
25934         reasons for disallowing fully-masked loops.  Make the final decision
25935         about whether use a fully-masked loop or not.
25936         (vect_estimate_min_profitable_iters): Do not assume that peeling
25937         for the number of iterations will be needed for fully-masked loops.
25938         (vectorizable_reduction): Disable fully-masked loops.
25939         (vectorizable_live_operation): Likewise.
25940         (vect_halve_mask_nunits): New function.
25941         (vect_double_mask_nunits): Likewise.
25942         (vect_record_loop_mask): Likewise.
25943         (vect_get_loop_mask): Likewise.
25944         (vect_transform_loop): Handle the case in which the final loop
25945         iteration might handle a partial vector.  Call vect_set_loop_condition
25946         instead of slpeel_make_loop_iterate_ntimes.
25947         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
25948         (check_load_store_masking): New function.
25949         (prepare_load_store_mask): Likewise.
25950         (vectorizable_store): Handle fully-masked loops.
25951         (vectorizable_load): Likewise.
25952         (supportable_widening_operation): Use vect_halve_mask_nunits for
25953         booleans.
25954         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
25955         (vect_gen_while): New function.
25956         * config/aarch64/aarch64.md (umax<mode>3): New expander.
25957         (aarch64_uqdec<mode>): New insn.
25959 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25960             Alan Hayward  <alan.hayward@arm.com>
25961             David Sherwood  <david.sherwood@arm.com>
25963         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
25964         (reduc_xor_scal_optab): New optabs.
25965         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
25966         (reduc_xor_scal_@var{m}): Document.
25967         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
25968         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
25969         internal functions.
25970         * fold-const-call.c (fold_const_call): Handle them.
25971         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
25972         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
25973         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
25974         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
25975         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
25976         (UNSPEC_XORV): New unspecs.
25977         (optab): Add entries for them.
25978         (BITWISEV): New int iterator.
25979         (bit_reduc_op): New int attributes.
25981 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
25982             Alan Hayward  <alan.hayward@arm.com>
25983             David Sherwood  <david.sherwood@arm.com>
25985         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
25986         * internal-fn.def (VEC_SHL_INSERT): New internal function.
25987         * optabs.def (vec_shl_insert_optab): New optab.
25988         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
25989         (duplicate_and_interleave): Likewise.
25990         * tree-vect-loop.c: Include internal-fn.h.
25991         (neutral_op_for_slp_reduction): New function, split out from
25992         get_initial_defs_for_reduction.
25993         (get_initial_def_for_reduction): Handle option 2 for variable-length
25994         vectors by loading the neutral value into a vector and then shifting
25995         the initial value into element 0.
25996         (get_initial_defs_for_reduction): Replace the code argument with
25997         the neutral value calculated by neutral_op_for_slp_reduction.
25998         Use gimple_build_vector for constant-length vectors.
25999         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
26000         but the first group_size elements have a neutral value.
26001         Use duplicate_and_interleave otherwise.
26002         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
26003         Update call to get_initial_defs_for_reduction.  Handle SLP
26004         reductions for variable-length vectors by creating one vector
26005         result for each scalar result, with the elements associated
26006         with other scalar results stubbed out with the neutral value.
26007         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
26008         Require IFN_VEC_SHL_INSERT for double reductions on
26009         variable-length vectors, or SLP reductions that have
26010         a neutral value.  Require can_duplicate_and_interleave_p
26011         support for variable-length unchained SLP reductions if there
26012         is no neutral value, such as for MIN/MAX reductions.  Also require
26013         the number of vector elements to be a multiple of the number of
26014         SLP statements when doing variable-length unchained SLP reductions.
26015         Update call to vect_create_epilog_for_reduction.
26016         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
26017         and remove initial values.
26018         (duplicate_and_interleave): Make public.
26019         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
26020         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
26022 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26023             Alan Hayward  <alan.hayward@arm.com>
26024             David Sherwood  <david.sherwood@arm.com>
26026         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
26027         (can_duplicate_and_interleave_p): New function.
26028         (vect_get_and_check_slp_defs): Take the vector of statements
26029         rather than just the current one.  Remove excess parentheses.
26030         Restriction rejectinon of vect_constant_def and vect_external_def
26031         for variable-length vectors to boolean types, or types for which
26032         can_duplicate_and_interleave_p is false.
26033         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
26034         (duplicate_and_interleave): New function.
26035         (vect_get_constant_vectors): Use gimple_build_vector for
26036         constant-length vectors and suitable variable-length constant
26037         vectors.  Use duplicate_and_interleave for other variable-length
26038         vectors.  Don't defer the update when inserting new statements.
26040 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26041             Alan Hayward  <alan.hayward@arm.com>
26042             David Sherwood  <david.sherwood@arm.com>
26044         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
26045         min_profitable_iters doesn't go negative.
26047 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26048             Alan Hayward  <alan.hayward@arm.com>
26049             David Sherwood  <david.sherwood@arm.com>
26051         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
26052         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
26053         * optabs.def (vec_mask_load_lanes_optab): New optab.
26054         (vec_mask_store_lanes_optab): Likewise.
26055         * internal-fn.def (MASK_LOAD_LANES): New internal function.
26056         (MASK_STORE_LANES): Likewise.
26057         * internal-fn.c (mask_load_lanes_direct): New macro.
26058         (mask_store_lanes_direct): Likewise.
26059         (expand_mask_load_optab_fn): Handle masked operations.
26060         (expand_mask_load_lanes_optab_fn): New macro.
26061         (expand_mask_store_optab_fn): Handle masked operations.
26062         (expand_mask_store_lanes_optab_fn): New macro.
26063         (direct_mask_load_lanes_optab_supported_p): Likewise.
26064         (direct_mask_store_lanes_optab_supported_p): Likewise.
26065         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
26066         parameter.
26067         (vect_load_lanes_supported): Likewise.
26068         * tree-vect-data-refs.c (strip_conversion): New function.
26069         (can_group_stmts_p): Likewise.
26070         (vect_analyze_data_ref_accesses): Use it instead of checking
26071         for a pair of assignments.
26072         (vect_store_lanes_supported): Take a masked_p parameter.
26073         (vect_load_lanes_supported): Likewise.
26074         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
26075         vect_store_lanes_supported and vect_load_lanes_supported.
26076         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
26077         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
26078         parameter.  Don't allow gaps for masked accesses.
26079         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
26080         and vect_load_lanes_supported.
26081         (get_load_store_type): Take a masked_p parameter and update
26082         call to get_group_load_store_type.
26083         (vectorizable_store): Update call to get_load_store_type.
26084         Handle IFN_MASK_STORE_LANES.
26085         (vectorizable_load): Update call to get_load_store_type.
26086         Handle IFN_MASK_LOAD_LANES.
26088 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26089             Alan Hayward  <alan.hayward@arm.com>
26090             David Sherwood  <david.sherwood@arm.com>
26092         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
26093         modes for SVE.
26094         * config/aarch64/aarch64-protos.h
26095         (aarch64_sve_struct_memory_operand_p): Declare.
26096         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
26097         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
26098         (VPRED, vpred): Handle SVE structure modes.
26099         * config/aarch64/constraints.md (Utx): New constraint.
26100         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
26101         (aarch64_sve_struct_nonimmediate_operand): New predicates.
26102         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
26103         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
26104         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
26105         structure modes.  Split into pieces after RA.
26106         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
26107         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
26108         New patterns.
26109         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
26110         SVE structure modes.
26111         (aarch64_classify_address): Likewise.
26112         (sizetochar): Move earlier in file.
26113         (aarch64_print_operand): Handle SVE register lists.
26114         (aarch64_array_mode): New function.
26115         (aarch64_sve_struct_memory_operand_p): Likewise.
26116         (TARGET_ARRAY_MODE): Redefine.
26118 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26119             Alan Hayward  <alan.hayward@arm.com>
26120             David Sherwood  <david.sherwood@arm.com>
26122         * target.def (array_mode): New target hook.
26123         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
26124         * doc/tm.texi: Regenerate.
26125         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
26126         * hooks.c (hook_optmode_mode_uhwi_none): New function.
26127         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
26128         targetm.array_mode.
26129         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
26130         type sizes.
26132 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26133             Alan Hayward  <alan.hayward@arm.com>
26134             David Sherwood  <david.sherwood@arm.com>
26136         * fold-const.c (fold_binary_loc): Check the argument types
26137         rather than the result type when testing for a vector operation.
26139 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26141         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
26142         * doc/tm.texi: Regenerate.
26144 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26145             Alan Hayward  <alan.hayward@arm.com>
26146             David Sherwood  <david.sherwood@arm.com>
26148         * doc/invoke.texi (-msve-vector-bits=): Document new option.
26149         (sve): Document new AArch64 extension.
26150         * doc/md.texi (w): Extend the description of the AArch64
26151         constraint to include SVE vectors.
26152         (Upl, Upa): Document new AArch64 predicate constraints.
26153         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
26154         enum.
26155         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
26156         (msve-vector-bits=): New option.
26157         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
26158         SVE when these are disabled.
26159         (sve): New extension.
26160         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
26161         modes.  Adjust their number of units based on aarch64_sve_vg.
26162         (MAX_BITSIZE_MODE_ANY_MODE): Define.
26163         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
26164         aarch64_addr_query_type.
26165         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
26166         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
26167         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
26168         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
26169         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
26170         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
26171         (aarch64_simd_imm_zero_p): Delete.
26172         (aarch64_check_zero_based_sve_index_immediate): Declare.
26173         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26174         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
26175         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
26176         (aarch64_sve_float_mul_immediate_p): Likewise.
26177         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
26178         rather than an rtx.
26179         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
26180         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
26181         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
26182         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
26183         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
26184         (aarch64_regmode_natural_size): Likewise.
26185         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
26186         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
26187         left one place.
26188         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
26189         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
26190         for VG and the SVE predicate registers.
26191         (V_ALIASES): Add a "z"-prefixed alias.
26192         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
26193         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
26194         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
26195         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
26196         (REG_CLASS_NAMES): Add entries for them.
26197         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
26198         and the predicate registers.
26199         (aarch64_sve_vg): Declare.
26200         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
26201         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
26202         (REGMODE_NATURAL_SIZE): Define.
26203         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
26204         SVE macros.
26205         * config/aarch64/aarch64.c: Include cfgrtl.h.
26206         (simd_immediate_info): Add a constructor for series vectors,
26207         and an associated step field.
26208         (aarch64_sve_vg): New variable.
26209         (aarch64_dbx_register_number): Handle VG and the predicate registers.
26210         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
26211         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
26212         (VEC_ANY_DATA, VEC_STRUCT): New constants.
26213         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
26214         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
26215         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
26216         (aarch64_get_mask_mode): New functions.
26217         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
26218         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
26219         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
26220         predicate modes and predicate registers.  Explicitly restrict
26221         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
26222         to store a vector mode if it is recognized by
26223         aarch64_classify_vector_mode.
26224         (aarch64_regmode_natural_size): New function.
26225         (aarch64_hard_regno_caller_save_mode): Return the original mode
26226         for predicates.
26227         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
26228         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
26229         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
26230         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
26231         functions.
26232         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
26233         does not overlap dest if the function is frame-related.  Handle
26234         SVE constants.
26235         (aarch64_split_add_offset): New function.
26236         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
26237         them aarch64_add_offset.
26238         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
26239         and update call to aarch64_sub_sp.
26240         (aarch64_add_cfa_expression): New function.
26241         (aarch64_expand_prologue): Pass extra temporary registers to the
26242         functions above.  Handle the case in which we need to emit new
26243         DW_CFA_expressions for registers that were originally saved
26244         relative to the stack pointer, but now have to be expressed
26245         relative to the frame pointer.
26246         (aarch64_output_mi_thunk): Pass extra temporary registers to the
26247         functions above.
26248         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
26249         IP0 and IP1 values for SVE frames.
26250         (aarch64_expand_vec_series): New function.
26251         (aarch64_expand_sve_widened_duplicate): Likewise.
26252         (aarch64_expand_sve_const_vector): Likewise.
26253         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
26254         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
26255         into the register, rather than emitting a SET directly.
26256         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
26257         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
26258         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
26259         (offset_9bit_signed_scaled_p): New functions.
26260         (aarch64_replicate_bitmask_imm): New function.
26261         (aarch64_bitmask_imm): Use it.
26262         (aarch64_cannot_force_const_mem): Reject expressions involving
26263         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
26264         (aarch64_classify_index): Handle SVE indices, by requiring
26265         a plain register index with a scale that matches the element size.
26266         (aarch64_classify_address): Handle SVE addresses.  Assert that
26267         the mode of the address is VOIDmode or an integer mode.
26268         Update call to aarch64_classify_symbol.
26269         (aarch64_classify_symbolic_expression): Update call to
26270         aarch64_classify_symbol.
26271         (aarch64_const_vec_all_in_range_p): New function.
26272         (aarch64_print_vector_float_operand): Likewise.
26273         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
26274         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
26275         and the FP immediates 1.0 and 0.5.
26276         (aarch64_print_address_internal): Handle SVE addresses.
26277         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
26278         (aarch64_regno_regclass): Handle predicate registers.
26279         (aarch64_secondary_reload): Handle big-endian reloads of SVE
26280         data modes.
26281         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
26282         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
26283         (aarch64_convert_sve_vector_bits): New function.
26284         (aarch64_override_options): Use it to handle -msve-vector-bits=.
26285         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
26286         rather than an rtx.
26287         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
26288         Handle SVE vector and predicate modes.  Accept VL-based constants
26289         that need only one temporary register, and VL offsets that require
26290         no temporary registers.
26291         (aarch64_conditional_register_usage): Mark the predicate registers
26292         as fixed if SVE isn't available.
26293         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
26294         Return true for SVE vector and predicate modes.
26295         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
26296         rather than an unsigned int.  Handle SVE modes.
26297         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
26298         SVE modes.
26299         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
26300         if SVE is enabled.
26301         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
26302         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
26303         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
26304         (aarch64_sve_float_mul_immediate_p): New functions.
26305         (aarch64_sve_valid_immediate): New function.
26306         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
26307         Explicitly reject structure modes.  Check for INDEX constants.
26308         Handle PTRUE and PFALSE constants.
26309         (aarch64_check_zero_based_sve_index_immediate): New function.
26310         (aarch64_simd_imm_zero_p): Delete.
26311         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
26312         vector modes.  Accept constants in the range of CNT[BHWD].
26313         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
26314         ask for an Advanced SIMD mode.
26315         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
26316         (aarch64_simd_vector_alignment): Handle SVE predicates.
26317         (aarch64_vectorize_preferred_vector_alignment): New function.
26318         (aarch64_simd_vector_alignment_reachable): Use it instead of
26319         the vector size.
26320         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
26321         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
26322         functions.
26323         (MAX_VECT_LEN): Delete.
26324         (expand_vec_perm_d): Add a vec_flags field.
26325         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
26326         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
26327         (aarch64_evpc_ext): Don't apply a big-endian lane correction
26328         for SVE modes.
26329         (aarch64_evpc_rev): Rename to...
26330         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
26331         (aarch64_evpc_rev_global): New function.
26332         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
26333         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
26334         MAX_VECT_LEN.
26335         (aarch64_evpc_sve_tbl): New function.
26336         (aarch64_expand_vec_perm_const_1): Update after rename of
26337         aarch64_evpc_rev.  Handle SVE permutes too, trying
26338         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
26339         than aarch64_evpc_tbl.
26340         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
26341         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
26342         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
26343         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
26344         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
26345         (aarch64_expand_sve_vcond): New functions.
26346         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
26347         of aarch64_vector_mode_p.
26348         (aarch64_dwarf_poly_indeterminate_value): New function.
26349         (aarch64_compute_pressure_classes): Likewise.
26350         (aarch64_can_change_mode_class): Likewise.
26351         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
26352         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
26353         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
26354         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
26355         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
26356         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
26357         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
26358         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
26359         constraints.
26360         (Dn, Dl, Dr): Accept const as well as const_vector.
26361         (Dz): Likewise.  Compare against CONST0_RTX.
26362         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
26363         of "vector" where appropriate.
26364         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
26365         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
26366         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
26367         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
26368         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
26369         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
26370         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
26371         (v_int_equiv): Extend to SVE modes.
26372         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
26373         mode attributes.
26374         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
26375         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
26376         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
26377         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
26378         (SVE_COND_FP_CMP): New int iterators.
26379         (perm_hilo): Handle the new unpack unspecs.
26380         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
26381         attributes.
26382         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
26383         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
26384         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
26385         (aarch64_equality_operator, aarch64_constant_vector_operand)
26386         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
26387         (aarch64_sve_nonimmediate_operand): Likewise.
26388         (aarch64_sve_general_operand): Likewise.
26389         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
26390         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
26391         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
26392         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
26393         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
26394         (aarch64_sve_float_arith_immediate): Likewise.
26395         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
26396         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
26397         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
26398         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
26399         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
26400         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
26401         (aarch64_sve_float_arith_operand): Likewise.
26402         (aarch64_sve_float_arith_with_sub_operand): Likewise.
26403         (aarch64_sve_float_mul_operand): Likewise.
26404         (aarch64_sve_vec_perm_operand): Likewise.
26405         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
26406         (aarch64_mov_operand): Accept const_poly_int and const_vector.
26407         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
26408         as well as const_vector.
26409         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
26410         in file.  Use CONST0_RTX and CONSTM1_RTX.
26411         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
26412         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
26413         Use aarch64_simd_imm_zero.
26414         * config/aarch64/aarch64-sve.md: New file.
26415         * config/aarch64/aarch64.md: Include it.
26416         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
26417         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
26418         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
26419         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
26420         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
26421         (sve): New attribute.
26422         (enabled): Disable instructions with the sve attribute unless
26423         TARGET_SVE.
26424         (movqi, movhi): Pass CONST_POLY_INT operaneds through
26425         aarch64_expand_mov_immediate.
26426         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
26427         CNT[BHSD] immediates.
26428         (movti): Split CONST_POLY_INT moves into two halves.
26429         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
26430         Split additions that need a temporary here if the destination
26431         is the stack pointer.
26432         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
26433         (*add<mode>3_poly_1): New instruction.
26434         (set_clobber_cc): New expander.
26436 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26438         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
26439         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
26440         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
26441         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
26442         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
26443         Change innermode from fixed_mode_size to machine_mode.
26444         (simplify_subreg): Update call accordingly.  Handle a constant-sized
26445         subreg of a variable-length CONST_VECTOR.
26447 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
26448             Alan Hayward  <alan.hayward@arm.com>
26449             David Sherwood  <david.sherwood@arm.com>
26451         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
26452         (add_offset_to_base): New function, split out from...
26453         (create_mem_ref): ...here.  When handling a scale other than 1,
26454         check first whether the address is valid without the offset.
26455         Add it into the base if so, leaving the index and scale as-is.
26457 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
26459         PR c++/83778
26460         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
26461         fold_for_warn before checking if arg2 is INTEGER_CST.
26463 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
26465         * config/rs6000/predicates.md (load_multiple_operation): Delete.
26466         (store_multiple_operation): Delete.
26467         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
26468         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
26469         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
26470         guarded by TARGET_STRING.
26471         (rs6000_output_load_multiple): Delete.
26472         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
26473         OPTION_MASK_STRING / TARGET_STRING handling.
26474         (print_operand) <'N', 'O'>: Add comment that these are unused now.
26475         (const rs6000_opt_masks) <"string">: Change mask to 0.
26476         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
26477         (MASK_STRING): Delete.
26478         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
26479         parts.  Simplify.
26480         (load_multiple): Delete.
26481         (*ldmsi8): Delete.
26482         (*ldmsi7): Delete.
26483         (*ldmsi6): Delete.
26484         (*ldmsi5): Delete.
26485         (*ldmsi4): Delete.
26486         (*ldmsi3): Delete.
26487         (store_multiple): Delete.
26488         (*stmsi8): Delete.
26489         (*stmsi7): Delete.
26490         (*stmsi6): Delete.
26491         (*stmsi5): Delete.
26492         (*stmsi4): Delete.
26493         (*stmsi3): Delete.
26494         (movmemsi_8reg): Delete.
26495         (corresponding unnamed define_insn): Delete.
26496         (movmemsi_6reg): Delete.
26497         (corresponding unnamed define_insn): Delete.
26498         (movmemsi_4reg): Delete.
26499         (corresponding unnamed define_insn): Delete.
26500         (movmemsi_2reg): Delete.
26501         (corresponding unnamed define_insn): Delete.
26502         (movmemsi_1reg): Delete.
26503         (corresponding unnamed define_insn): Delete.
26504         * config/rs6000/rs6000.opt (mno-string): New.
26505         (mstring): Replace by deprecation warning stub.
26506         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
26508 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
26510         * regrename.c (regrename_do_replace): If replacing the same
26511         reg multiple times, try to reuse last created gen_raw_REG.
26513         PR debug/81155
26514         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
26515         main to workaround a bug in GDB.
26517 2018-01-12  Tom de Vries  <tom@codesourcery.com>
26519         PR target/83737
26520         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
26522 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
26524         PR rtl-optimization/80481
26525         * ira-color.c (get_cap_member): New function.
26526         (allocnos_conflict_by_live_ranges_p): Use it.
26527         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
26528         (setup_slot_coalesced_allocno_live_ranges): Ditto.
26530 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
26532         PR target/83628
26533         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
26534         (*saddl_se_1): Ditto.
26535         (*ssubsi_1): Ditto.
26536         (*ssubl_se_1): Ditto.
26538 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
26540         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
26541         rather than wi::to_widest for DR_INITs.
26542         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
26543         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
26544         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
26545         INTEGER_CSTs.
26546         (vect_analyze_group_access_1): Note that here.
26548 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
26550         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
26551         polynomial type sizes.
26553 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
26555         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
26556         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
26557         (gimple_add_tmp_var): Likewise.
26559 2018-01-12  Martin Liska  <mliska@suse.cz>
26561         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
26562         (gimple_alloc_sizes): Likewise.
26563         (dump_gimple_statistics): Use PRIu64 in printf format.
26564         * gimple.h: Change uint64_t to int.
26566 2018-01-12  Martin Liska  <mliska@suse.cz>
26568         * tree-core.h: Use uint64_t instead of int.
26569         * tree.c (tree_node_counts): Likewise.
26570         (tree_node_sizes): Likewise.
26571         (dump_tree_statistics): Use PRIu64 in printf format.
26573 2018-01-12  Martin Liska  <mliska@suse.cz>
26575         * Makefile.in: As qsort_chk is implemented in vec.c, add
26576         vec.o to linkage of gencfn-macros.
26577         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
26578         passing the info to record_node_allocation_statistics.
26579         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
26580         and pass the info.
26581         * ggc-common.c (struct ggc_usage): Add operator== and use
26582         it in operator< and compare function.
26583         * mem-stats.h (struct mem_usage): Likewise.
26584         * vec.c (struct vec_usage): Remove operator< and compare
26585         function. Can be simply inherited.
26587 2018-01-12  Martin Jambor  <mjambor@suse.cz>
26589         PR target/81616
26590         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
26591         * tree-ssa-math-opts.c: Include domwalk.h.
26592         (convert_mult_to_fma_1): New function.
26593         (fma_transformation_info): New type.
26594         (fma_deferring_state): Likewise.
26595         (cancel_fma_deferring): New function.
26596         (result_of_phi): Likewise.
26597         (last_fma_candidate_feeds_initial_phi): Likewise.
26598         (convert_mult_to_fma): Added deferring logic, split actual
26599         transformation to convert_mult_to_fma_1.
26600         (math_opts_dom_walker): New type.
26601         (math_opts_dom_walker::after_dom_children): New method, body moved
26602         here from pass_optimize_widening_mul::execute, added deferring logic
26603         bits.
26604         (pass_optimize_widening_mul::execute): Moved most of code to
26605         math_opts_dom_walker::after_dom_children.
26606         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
26607         * config/i386/i386.c (ix86_option_override_internal): Added
26608         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
26610 2018-01-12  Richard Biener  <rguenther@suse.de>
26612         PR debug/83157
26613         * dwarf2out.c (gen_variable_die): Do not reset old_die for
26614         inline instance vars.
26616 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
26618         PR target/81819
26619         * config/rx/rx.c (rx_is_restricted_memory_address):
26620         Handle SUBREG case.
26622 2018-01-12  Richard Biener  <rguenther@suse.de>
26624         PR tree-optimization/80846
26625         * target.def (split_reduction): New target hook.
26626         * targhooks.c (default_split_reduction): New function.
26627         * targhooks.h (default_split_reduction): Declare.
26628         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
26629         target requests first reduce vectors by combining low and high
26630         parts.
26631         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
26632         (get_vectype_for_scalar_type_and_size): Export.
26633         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
26634         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
26635         * doc/tm.texi: Regenerate.
26636         * config/i386/i386.c (ix86_split_reduction): Implement
26637         TARGET_VECTORIZE_SPLIT_REDUCTION.
26639 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26641         PR target/83368
26642         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
26643         in PIC mode except for TARGET_VXWORKS_RTP.
26644         * config/sparc/sparc.c: Include cfgrtl.h.
26645         (TARGET_INIT_PIC_REG): Define.
26646         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
26647         (sparc_pic_register_p): New predicate.
26648         (sparc_legitimate_address_p): Use it.
26649         (sparc_legitimize_pic_address): Likewise.
26650         (sparc_delegitimize_address): Likewise.
26651         (sparc_mode_dependent_address_p): Likewise.
26652         (gen_load_pcrel_sym): Remove 4th parameter.
26653         (load_got_register): Adjust call to above.  Remove obsolete stuff.
26654         (sparc_expand_prologue): Do not call load_got_register here.
26655         (sparc_flat_expand_prologue): Likewise.
26656         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
26657         (sparc_use_pseudo_pic_reg): New function.
26658         (sparc_init_pic_reg): Likewise.
26659         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26660         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
26662 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
26664         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
26665         Add item for branch_cost.
26667 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26669         PR rtl-optimization/83565
26670         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
26671         not extend the result to a larger mode for rotate operations.
26672         (num_sign_bit_copies1): Likewise.
26674 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26676         PR target/40411
26677         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
26678         -symbolic.
26679         Use values-Xc.o for -pedantic.
26680         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
26682 2018-01-12  Martin Liska  <mliska@suse.cz>
26684         PR ipa/83054
26685         * ipa-devirt.c (final_warning_record::grow_type_warnings):
26686         New function.
26687         (possible_polymorphic_call_targets): Use it.
26688         (ipa_devirt): Likewise.
26690 2018-01-12  Martin Liska  <mliska@suse.cz>
26692         * profile-count.h (enum profile_quality): Use 0 as invalid
26693         enum value of profile_quality.
26695 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
26697         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
26698         -mext-string options.
26700 2018-01-12  Richard Biener  <rguenther@suse.de>
26702         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
26703         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
26704         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
26705         Likewise.
26706         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
26708 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
26710         * configure.ac (--with-long-double-format): Add support for the
26711         configuration option to change the default long double format on
26712         PowerPC systems.
26713         * config.gcc (powerpc*-linux*-*): Likewise.
26714         * configure: Regenerate.
26715         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
26716         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
26717         used without modification.
26719 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26721         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
26722         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
26723         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
26724         MISC_BUILTIN_SPEC_BARRIER.
26725         (rs6000_init_builtins): Likewise.
26726         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
26727         enum value.
26728         (speculation_barrier): New define_insn.
26729         * doc/extend.texi: Document __builtin_speculation_barrier.
26731 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
26733         PR target/83203
26734         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
26735         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
26736         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
26737         iterators.
26738         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
26739         integral modes instead of "ss" and "sd".
26740         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
26741         vectors with 32-bit and 64-bit elements.
26742         (vecdupssescalarmodesuffix): New mode attribute.
26743         (vec_dup<mode>): Use it.
26745 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
26747         PR target/83330
26748         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
26749         frame if argument is passed on stack.
26751 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
26753         PR target/82682
26754         * ree.c (combine_reaching_defs): Optimize also
26755         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
26756         reg2=any_extend(exp); reg1=reg2;, formatting fix.
26758 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
26760         PR middle-end/83189
26761         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
26763 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
26765         PR middle-end/83718
26766         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
26767         after they are computed.
26769 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
26771         PR tree-optimization/83695
26772         * gimple-loop-linterchange.cc
26773         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
26774         reset cached scev information after interchange.
26775         (pass_linterchange::execute): Remove call to scev_reset_htab.
26777 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26779         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
26780         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
26781         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
26782         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
26783         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
26784         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
26785         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
26786         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
26787         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
26788         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
26789         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
26790         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
26791         (V_lane_reg): Likewise.
26792         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
26793         New define_expand.
26794         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
26795         (vfmal_lane_low<mode>_intrinsic,
26796         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
26797         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
26798         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
26799         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
26800         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
26801         vfmsl_lane_high<mode>_intrinsic): New define_insns.
26803 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26805         * config/arm/arm-cpus.in (fp16fml): New feature.
26806         (ALL_SIMD): Add fp16fml.
26807         (armv8.2-a): Add fp16fml as an option.
26808         (armv8.3-a): Likewise.
26809         (armv8.4-a): Add fp16fml as part of fp16.
26810         * config/arm/arm.h (TARGET_FP16FML): Define.
26811         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
26812         when appropriate.
26813         * config/arm/arm-modes.def (V2HF): Define.
26814         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
26815         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
26816         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
26817         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
26818         vfmsl_low, vfmsl_high): New set of builtins.
26819         * config/arm/iterators.md (PLUSMINUS): New code iterator.
26820         (vfml_op): New code attribute.
26821         (VFMLHALVES): New int iterator.
26822         (VFML, VFMLSEL): New mode attributes.
26823         (V_reg): Define mapping for V2HF.
26824         (V_hi, V_lo): New mode attributes.
26825         (VF_constraint): Likewise.
26826         (vfml_half, vfml_half_selector): New int attributes.
26827         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
26828         define_expand.
26829         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
26830         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
26831         New define_insn.
26832         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
26833         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
26834         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
26835         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
26836         documentation.
26837         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
26838         Document new effective target and option set.
26840 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26842         * config/arm/arm-cpus.in (armv8_4): New feature.
26843         (ARMv8_4a): New fgroup.
26844         (armv8.4-a): New arch.
26845         * config/arm/arm-tables.opt: Regenerate.
26846         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
26847         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
26848         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
26849         Add matching rules for -march=armv8.4-a and extensions.
26850         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
26852 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
26854         PR target/81821
26855         * config/rx/rx.md (BW): New mode attribute.
26856         (sync_lock_test_and_setsi): Add mode suffix to insn output.
26858 2018-01-11  Richard Biener  <rguenther@suse.de>
26860         PR tree-optimization/83435
26861         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
26862         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
26863         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
26865 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26866             Alan Hayward  <alan.hayward@arm.com>
26867             David Sherwood  <david.sherwood@arm.com>
26869         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
26870         field.
26871         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
26872         (aarch64_print_address_internal): Use it to check for a zero offset.
26874 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26875             Alan Hayward  <alan.hayward@arm.com>
26876             David Sherwood  <david.sherwood@arm.com>
26878         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
26879         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
26880         Return a poly_int64 rather than a HOST_WIDE_INT.
26881         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
26882         rather than a HOST_WIDE_INT.
26883         * config/aarch64/aarch64.h (aarch64_frame): Protect with
26884         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
26885         hard_fp_offset, frame_size, initial_adjust, callee_offset and
26886         final_offset from HOST_WIDE_INT to poly_int64.
26887         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
26888         to_constant when getting the number of units in an Advanced SIMD
26889         mode.
26890         (aarch64_builtin_vectorized_function): Check for a constant number
26891         of units.
26892         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
26893         GET_MODE_SIZE.
26894         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
26895         attribute instead of GET_MODE_NUNITS.
26896         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
26897         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
26898         GET_MODE_SIZE for fixed-size registers.
26899         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
26900         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
26901         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
26902         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
26903         (aarch64_print_operand, aarch64_print_address_internal)
26904         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
26905         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
26906         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
26907         Handle polynomial GET_MODE_SIZE.
26908         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
26909         wider than SImode without modification.
26910         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
26911         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
26912         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
26913         passing and returning SVE modes.
26914         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
26915         rather than GEN_INT.
26916         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
26917         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
26918         (aarch64_allocate_and_probe_stack_space): Likewise.
26919         (aarch64_layout_frame): Cope with polynomial offsets.
26920         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
26921         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
26922         polynomial offsets.
26923         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
26924         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
26925         poly_int64 rather than a HOST_WIDE_INT.
26926         (aarch64_get_separate_components, aarch64_process_components)
26927         (aarch64_expand_prologue, aarch64_expand_epilogue)
26928         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
26929         (aarch64_anchor_offset): New function, split out from...
26930         (aarch64_legitimize_address): ...here.
26931         (aarch64_builtin_vectorization_cost): Handle polynomial
26932         TYPE_VECTOR_SUBPARTS.
26933         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
26934         GET_MODE_NUNITS.
26935         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
26936         number of elements from the PARALLEL rather than the mode.
26937         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
26938         rather than GET_MODE_BITSIZE.
26939         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
26940         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
26941         (aarch64_expand_vec_perm_const_1): Handle polynomial
26942         d->perm.length () and d->perm elements.
26943         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
26944         Apply to_constant to d->perm elements.
26945         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
26946         polynomial CONST_VECTOR_NUNITS.
26947         (aarch64_move_pointer): Take amount as a poly_int64 rather
26948         than an int.
26949         (aarch64_progress_pointer): Avoid temporary variable.
26950         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
26951         the mode attribute instead of GET_MODE.
26953 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26954             Alan Hayward  <alan.hayward@arm.com>
26955             David Sherwood  <david.sherwood@arm.com>
26957         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
26958         x exists before using it.
26959         (aarch64_add_constant_internal): Rename to...
26960         (aarch64_add_offset_1): ...this.  Replace regnum with separate
26961         src and dest rtxes.  Handle the case in which they're different,
26962         including when the offset is zero.  Replace scratchreg with an rtx.
26963         Use 2 additions if there is no spare register into which we can
26964         move a 16-bit constant.
26965         (aarch64_add_constant): Delete.
26966         (aarch64_add_offset): Replace reg with separate src and dest
26967         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
26968         Use aarch64_add_offset_1.
26969         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
26970         an rtx rather than an int.  Take the delta as a poly_int64
26971         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
26972         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
26973         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
26974         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
26975         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
26976         and aarch64_add_sp.
26977         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
26978         aarch64_add_constant.
26980 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26982         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
26983         Use scalar_float_mode.
26985 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
26987         * config/aarch64/aarch64-simd.md
26988         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
26989         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
26990         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
26991         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
26992         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
26993         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
26994         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
26995         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
26996         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
26997         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
26999 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27001         PR target/83514
27002         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
27003         targ_options->x_arm_arch_string is non NULL.
27005 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
27007         * config/aarch64/aarch64.h
27008         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
27010 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
27012         PR target/82096
27013         * expmed.c (emit_store_flag_force): Swap if const op0
27014         and change VOIDmode to mode of op0.
27016 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
27018         PR rtl-optimization/83761
27019         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
27020         than bytes to mode_for_size.
27022 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27024         PR middle-end/83189
27025         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
27026         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
27027         profile.
27029 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27031         PR middle-end/83575
27032         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
27033         when in layout mode.
27034         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
27035         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
27036         partition fixup.
27038 2018-01-10  Michael Collison  <michael.collison@arm.com>
27040         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
27041         * config/aarch64/aarch64-option-extension.def: Add
27042         AARCH64_OPT_EXTENSION of 'fp16fml'.
27043         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27044         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
27045         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
27046         * config/aarch64/constraints.md (Ui7): New constraint.
27047         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
27048         (VFMLA_SEL_W): Ditto.
27049         (f16quad): Ditto.
27050         (f16mac1): Ditto.
27051         (VFMLA16_LOW): New int iterator.
27052         (VFMLA16_HIGH): Ditto.
27053         (UNSPEC_FMLAL): New unspec.
27054         (UNSPEC_FMLSL): Ditto.
27055         (UNSPEC_FMLAL2): Ditto.
27056         (UNSPEC_FMLSL2): Ditto.
27057         (f16mac): New code attribute.
27058         * config/aarch64/aarch64-simd-builtins.def
27059         (aarch64_fmlal_lowv2sf): Ditto.
27060         (aarch64_fmlsl_lowv2sf): Ditto.
27061         (aarch64_fmlalq_lowv4sf): Ditto.
27062         (aarch64_fmlslq_lowv4sf): Ditto.
27063         (aarch64_fmlal_highv2sf): Ditto.
27064         (aarch64_fmlsl_highv2sf): Ditto.
27065         (aarch64_fmlalq_highv4sf): Ditto.
27066         (aarch64_fmlslq_highv4sf): Ditto.
27067         (aarch64_fmlal_lane_lowv2sf): Ditto.
27068         (aarch64_fmlsl_lane_lowv2sf): Ditto.
27069         (aarch64_fmlal_laneq_lowv2sf): Ditto.
27070         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
27071         (aarch64_fmlalq_lane_lowv4sf): Ditto.
27072         (aarch64_fmlsl_lane_lowv4sf): Ditto.
27073         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
27074         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
27075         (aarch64_fmlal_lane_highv2sf): Ditto.
27076         (aarch64_fmlsl_lane_highv2sf): Ditto.
27077         (aarch64_fmlal_laneq_highv2sf): Ditto.
27078         (aarch64_fmlsl_laneq_highv2sf): Ditto.
27079         (aarch64_fmlalq_lane_highv4sf): Ditto.
27080         (aarch64_fmlsl_lane_highv4sf): Ditto.
27081         (aarch64_fmlalq_laneq_highv4sf): Ditto.
27082         (aarch64_fmlsl_laneq_highv4sf): Ditto.
27083         * config/aarch64/aarch64-simd.md:
27084         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
27085         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27086         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
27087         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
27088         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
27089         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
27090         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
27091         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
27092         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
27093         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
27094         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
27095         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
27096         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
27097         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
27098         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
27099         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
27100         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
27101         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
27102         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
27103         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
27104         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
27105         (vfmlsl_low_u32): Ditto.
27106         (vfmlalq_low_u32): Ditto.
27107         (vfmlslq_low_u32): Ditto.
27108         (vfmlal_high_u32): Ditto.
27109         (vfmlsl_high_u32): Ditto.
27110         (vfmlalq_high_u32): Ditto.
27111         (vfmlslq_high_u32): Ditto.
27112         (vfmlal_lane_low_u32): Ditto.
27113         (vfmlsl_lane_low_u32): Ditto.
27114         (vfmlal_laneq_low_u32): Ditto.
27115         (vfmlsl_laneq_low_u32): Ditto.
27116         (vfmlalq_lane_low_u32): Ditto.
27117         (vfmlslq_lane_low_u32): Ditto.
27118         (vfmlalq_laneq_low_u32): Ditto.
27119         (vfmlslq_laneq_low_u32): Ditto.
27120         (vfmlal_lane_high_u32): Ditto.
27121         (vfmlsl_lane_high_u32): Ditto.
27122         (vfmlal_laneq_high_u32): Ditto.
27123         (vfmlsl_laneq_high_u32): Ditto.
27124         (vfmlalq_lane_high_u32): Ditto.
27125         (vfmlslq_lane_high_u32): Ditto.
27126         (vfmlalq_laneq_high_u32): Ditto.
27127         (vfmlslq_laneq_high_u32): Ditto.
27128         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
27129         (AARCH64_FL_FOR_ARCH8_4): New.
27130         (AARCH64_ISA_F16FML): New ISA flag.
27131         (TARGET_F16FML): New feature flag for fp16fml.
27132         (doc/invoke.texi): Document new fp16fml option.
27134 2018-01-10  Michael Collison  <michael.collison@arm.com>
27136         * config/aarch64/aarch64-builtins.c:
27137         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
27138         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27139         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
27140         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
27141         (AARCH64_ISA_SHA3): New ISA flag.
27142         (TARGET_SHA3): New feature flag for sha3.
27143         * config/aarch64/iterators.md (sha512_op): New int attribute.
27144         (CRYPTO_SHA512): New int iterator.
27145         (UNSPEC_SHA512H): New unspec.
27146         (UNSPEC_SHA512H2): Ditto.
27147         (UNSPEC_SHA512SU0): Ditto.
27148         (UNSPEC_SHA512SU1): Ditto.
27149         * config/aarch64/aarch64-simd-builtins.def
27150         (aarch64_crypto_sha512hqv2di): New builtin.
27151         (aarch64_crypto_sha512h2qv2di): Ditto.
27152         (aarch64_crypto_sha512su0qv2di): Ditto.
27153         (aarch64_crypto_sha512su1qv2di): Ditto.
27154         (aarch64_eor3qv8hi): Ditto.
27155         (aarch64_rax1qv2di): Ditto.
27156         (aarch64_xarqv2di): Ditto.
27157         (aarch64_bcaxqv8hi): Ditto.
27158         * config/aarch64/aarch64-simd.md:
27159         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
27160         (aarch64_crypto_sha512su0qv2di): Ditto.
27161         (aarch64_crypto_sha512su1qv2di): Ditto.
27162         (aarch64_eor3qv8hi): Ditto.
27163         (aarch64_rax1qv2di): Ditto.
27164         (aarch64_xarqv2di): Ditto.
27165         (aarch64_bcaxqv8hi): Ditto.
27166         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
27167         (vsha512h2q_u64): Ditto.
27168         (vsha512su0q_u64): Ditto.
27169         (vsha512su1q_u64): Ditto.
27170         (veor3q_u16): Ditto.
27171         (vrax1q_u64): Ditto.
27172         (vxarq_u64): Ditto.
27173         (vbcaxq_u16): Ditto.
27174         * config/arm/types.md (crypto_sha512): New type attribute.
27175         (crypto_sha3): Ditto.
27176         (doc/invoke.texi): Document new sha3 option.
27178 2018-01-10  Michael Collison  <michael.collison@arm.com>
27180         * config/aarch64/aarch64-builtins.c:
27181         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
27182         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27183         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
27184         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
27185         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
27186         (AARCH64_ISA_SM4): New ISA flag.
27187         (TARGET_SM4): New feature flag for sm4.
27188         * config/aarch64/aarch64-simd-builtins.def
27189         (aarch64_sm3ss1qv4si): Ditto.
27190         (aarch64_sm3tt1aq4si): Ditto.
27191         (aarch64_sm3tt1bq4si): Ditto.
27192         (aarch64_sm3tt2aq4si): Ditto.
27193         (aarch64_sm3tt2bq4si): Ditto.
27194         (aarch64_sm3partw1qv4si): Ditto.
27195         (aarch64_sm3partw2qv4si): Ditto.
27196         (aarch64_sm4eqv4si): Ditto.
27197         (aarch64_sm4ekeyqv4si): Ditto.
27198         * config/aarch64/aarch64-simd.md:
27199         (aarch64_sm3ss1qv4si): Ditto.
27200         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
27201         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
27202         (aarch64_sm4eqv4si): Ditto.
27203         (aarch64_sm4ekeyqv4si): Ditto.
27204         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
27205         (sm3part_op): Ditto.
27206         (CRYPTO_SM3TT): Ditto.
27207         (CRYPTO_SM3PART): Ditto.
27208         (UNSPEC_SM3SS1): New unspec.
27209         (UNSPEC_SM3TT1A): Ditto.
27210         (UNSPEC_SM3TT1B): Ditto.
27211         (UNSPEC_SM3TT2A): Ditto.
27212         (UNSPEC_SM3TT2B): Ditto.
27213         (UNSPEC_SM3PARTW1): Ditto.
27214         (UNSPEC_SM3PARTW2): Ditto.
27215         (UNSPEC_SM4E): Ditto.
27216         (UNSPEC_SM4EKEY): Ditto.
27217         * config/aarch64/constraints.md (Ui2): New constraint.
27218         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
27219         * config/arm/types.md (crypto_sm3): New type attribute.
27220         (crypto_sm4): Ditto.
27221         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
27222         (vsm3tt1aq_u32): Ditto.
27223         (vsm3tt1bq_u32): Ditto.
27224         (vsm3tt2aq_u32): Ditto.
27225         (vsm3tt2bq_u32): Ditto.
27226         (vsm3partw1q_u32): Ditto.
27227         (vsm3partw2q_u32): Ditto.
27228         (vsm4eq_u32): Ditto.
27229         (vsm4ekeyq_u32): Ditto.
27230         (doc/invoke.texi): Document new sm4 option.
27232 2018-01-10  Michael Collison  <michael.collison@arm.com>
27234         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
27235         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
27236         (AARCH64_FL_FOR_ARCH8_4): New.
27237         (AARCH64_FL_V8_4): New flag.
27238         (doc/invoke.texi): Document new armv8.4-a option.
27240 2018-01-10  Michael Collison  <michael.collison@arm.com>
27242         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
27243         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
27244         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
27245         * config/aarch64/aarch64-option-extension.def: Add
27246         AARCH64_OPT_EXTENSION of 'sha2'.
27247         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
27248         (crypto): Disable sha2 and aes if crypto disabled.
27249         (crypto): Enable aes and sha2 if enabled.
27250         (simd): Disable sha2 and aes if simd disabled.
27251         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
27252         New flags.
27253         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
27254         (TARGET_SHA2): New feature flag for sha2.
27255         (TARGET_AES): New feature flag for aes.
27256         * config/aarch64/aarch64-simd.md:
27257         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
27258         conditional on TARGET_AES.
27259         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
27260         (aarch64_crypto_sha1hsi): Make pattern conditional
27261         on TARGET_SHA2.
27262         (aarch64_crypto_sha1hv4si): Ditto.
27263         (aarch64_be_crypto_sha1hv4si): Ditto.
27264         (aarch64_crypto_sha1su1v4si): Ditto.
27265         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
27266         (aarch64_crypto_sha1su0v4si): Ditto.
27267         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
27268         (aarch64_crypto_sha256su0v4si): Ditto.
27269         (aarch64_crypto_sha256su1v4si): Ditto.
27270         (doc/invoke.texi): Document new aes and sha2 options.
27272 2018-01-10  Martin Sebor  <msebor@redhat.com>
27274         PR tree-optimization/83781
27275         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
27276         as string arrays.
27278 2018-01-11  Martin Sebor  <msebor@gmail.com>
27279             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27281         PR tree-optimization/83501
27282         PR tree-optimization/81703
27284         * tree-ssa-strlen.c (get_string_cst): Rename...
27285         (get_string_len): ...to this.  Handle global constants.
27286         (handle_char_store): Adjust.
27288 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
27289             Jim Wilson  <jimw@sifive.com>
27291         * config/riscv/riscv-protos.h (riscv_output_return): New.
27292         * config/riscv/riscv.c (struct machine_function): New naked_p field.
27293         (riscv_attribute_table, riscv_output_return),
27294         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
27295         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
27296         (riscv_compute_frame_info): Only compute frame->mask if not a naked
27297         function.
27298         (riscv_expand_prologue): Add early return for naked function.
27299         (riscv_expand_epilogue): Likewise.
27300         (riscv_function_ok_for_sibcall): Return false for naked function.
27301         (riscv_set_current_function): New.
27302         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
27303         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
27304         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
27305         * doc/extend.texi (RISC-V Function Attributes): New.
27307 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
27309         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
27310         check for 128-bit long double before checking TCmode.
27311         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
27312         128-bit long doubles before checking TFmode or TCmode.
27313         (FLOAT128_IBM_P): Likewise.
27315 2018-01-10  Martin Sebor  <msebor@redhat.com>
27317         PR tree-optimization/83671
27318         * builtins.c (c_strlen): Unconditionally return zero for the empty
27319         string.
27320         Use -Warray-bounds for warnings.
27321         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
27322         for non-constant array indices with COMPONENT_REF, arrays of
27323         arrays, and pointers to arrays.
27324         (gimple_fold_builtin_strlen): Determine and set length range for
27325         non-constant character arrays.
27327 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
27329         PR middle-end/81897
27330         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
27331         empty blocks.
27333 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
27335         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
27337 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
27339         PR target/83399
27340         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
27341         VECTOR_MEM_ALTIVEC_OR_VSX_P.
27342         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
27343         indexed_or_indirect_operand predicate.
27344         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
27345         (*vsx_le_perm_load_v8hi): Likewise.
27346         (*vsx_le_perm_load_v16qi): Likewise.
27347         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
27348         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
27349         (*vsx_le_perm_store_v8hi): Likewise.
27350         (*vsx_le_perm_store_v16qi): Likewise.
27351         (eight unnamed splitters): Likewise.
27353 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
27355         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
27356         * config/rs6000/emmintrin.h: Likewise.
27357         * config/rs6000/mmintrin.h: Likewise.
27358         * config/rs6000/xmmintrin.h: Likewise.
27360 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
27362         PR c++/43486
27363         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
27364         "public_flag".
27365         * tree.c (tree_nop_conversion): Return true for location wrapper
27366         nodes.
27367         (maybe_wrap_with_location): New function.
27368         (selftest::check_strip_nops): New function.
27369         (selftest::test_location_wrappers): New function.
27370         (selftest::tree_c_tests): Call it.
27371         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
27372         (maybe_wrap_with_location): New decl.
27373         (EXPR_LOCATION_WRAPPER_P): New macro.
27374         (location_wrapper_p): New inline function.
27375         (tree_strip_any_location_wrapper): New inline function.
27377 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
27379         PR target/83735
27380         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
27381         stack_realign_offset for the largest alignment of stack slot
27382         actually used.
27383         (ix86_find_max_used_stack_alignment): New function.
27384         (ix86_finalize_stack_frame_flags): Use it.  Set
27385         max_used_stack_alignment if we don't realign stack.
27386         * config/i386/i386.h (machine_function): Add
27387         max_used_stack_alignment.
27389 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
27391         * config/arm/arm.opt (-mbranch-cost): New option.
27392         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
27393         account.
27395 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
27397         PR target/83629
27398         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
27399         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
27401 2018-01-10  Richard Biener  <rguenther@suse.de>
27403         PR debug/83765
27404         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
27405         early out so it also covers the case where we have a non-NULL
27406         origin.
27408 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
27410         PR tree-optimization/83753
27411         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
27412         for non-strided grouped accesses if the number of elements is 1.
27414 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
27416         PR target/81616
27417         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
27418         * i386.h (TARGET_USE_GATHER): Define.
27419         * x86-tune.def (X86_TUNE_USE_GATHER): New.
27421 2018-01-10  Martin Liska  <mliska@suse.cz>
27423         PR bootstrap/82831
27424         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
27425         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
27426         partitioning.
27427         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
27428         CLEANUP_NO_PARTITIONING is not set.
27430 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
27432         * doc/rtl.texi: Remove documentation of (const ...) wrappers
27433         for vectors, as a partial revert of r254296.
27434         * rtl.h (const_vec_p): Delete.
27435         (const_vec_duplicate_p): Don't test for vector CONSTs.
27436         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
27437         * expmed.c (make_tree): Likewise.
27439         Revert:
27440         * common.md (E, F): Use CONSTANT_P instead of checking for
27441         CONST_VECTOR.
27442         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
27443         checking for CONST_VECTOR.
27445 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
27447         PR middle-end/83575
27448         * predict.c (force_edge_cold): Handle in more sane way edges
27449         with no prediction.
27451 2018-01-09  Carl Love  <cel@us.ibm.com>
27453         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
27454         V4SI, V4SF types.
27455         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
27456         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
27457         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
27458         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
27459         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
27460         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
27461         * config/rs6000/rs6000-protos.h: Add extern defition for
27462         rs6000_generate_float2_double_code.
27463         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
27464         function.
27465         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
27466         (float2_v2df): Add define_expand.
27468 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
27470         PR target/83628
27471         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
27472         op_mode in the force_to_mode call.
27474 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
27476         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
27477         instead of checking each element individually.
27478         (aarch64_evpc_uzp): Likewise.
27479         (aarch64_evpc_zip): Likewise.
27480         (aarch64_evpc_ext): Likewise.
27481         (aarch64_evpc_rev): Likewise.
27482         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
27483         instead of checking each element individually.  Return true without
27484         generating rtl if
27485         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
27486         whether all selected elements come from the same input, instead of
27487         checking each element individually.  Remove calls to gen_rtx_REG,
27488         start_sequence and end_sequence and instead assert that no rtl is
27489         generated.
27491 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
27493         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
27494         order of HIGH and CONST checks.
27496 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
27498         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
27499         if the destination isn't an SSA_NAME.
27501 2018-01-09  Richard Biener  <rguenther@suse.de>
27503         PR tree-optimization/83668
27504         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
27505         move prologue...
27506         (canonicalize_loop_form): ... here, renamed from ...
27507         (canonicalize_loop_closed_ssa_form): ... this and amended to
27508         swap successor edges for loop exit blocks to make us use
27509         the RPO order we need for initial schedule generation.
27511 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
27513         PR tree-optimization/64811
27514         * match.pd: When optimizing comparisons with Inf, avoid
27515         introducing or losing exceptions from comparisons with NaN.
27517 2018-01-09  Martin Liska  <mliska@suse.cz>
27519         PR sanitizer/82517
27520         * asan.c (shadow_mem_size): Add gcc_assert.
27522 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
27524         Don't save registers in main().
27526         PR target/83738
27527         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
27528         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
27529         * config/avr/avr.c (avr_set_current_function): Don't error if
27530         naked, OS_task or OS_main are specified at the same time.
27531         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
27532         OS_main.
27533         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
27534         attribute.
27535         * common/config/avr/avr-common.c (avr_option_optimization_table):
27536         Switch on -mmain-is-OS_task for optimizing compilations.
27538 2018-01-09  Richard Biener  <rguenther@suse.de>
27540         PR tree-optimization/83572
27541         * graphite.c: Include cfganal.h.
27542         (graphite_transform_loops): Connect infinite loops to exit
27543         and remove fake edges at the end.
27545 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
27547         * ipa-inline.c (edge_badness): Revert accidental checkin.
27549 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
27551         PR ipa/80763
27552         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
27553         symbols; not inline clones.
27555 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
27557         PR target/83507
27558         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
27559         hard registers.  Formatting fixes.
27561         PR preprocessor/83722
27562         * gcc.c (try_generate_repro): Pass
27563         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
27564         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
27565         do_report_bug.
27567 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
27568             Kito Cheng  <kito.cheng@gmail.com>
27570         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
27571         (riscv_leaf_function_p): Delete.
27572         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
27574 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27576         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
27577         function.
27578         (do_ifelse): New function.
27579         (do_isel): New function.
27580         (do_sub3): New function.
27581         (do_add3): New function.
27582         (do_load_mask_compare): New function.
27583         (do_overlap_load_compare): New function.
27584         (expand_compare_loop): New function.
27585         (expand_block_compare): Call expand_compare_loop() when appropriate.
27586         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
27587         option description.
27588         (-mblock-compare-inline-loop-limit): New option.
27590 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27592         PR target/83677
27593         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
27594         Reverse order of second and third operands in first alternative.
27595         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
27596         of first and second elements in UNSPEC_VPERMR vector.
27597         (altivec_expand_vec_perm_le): Likewise.
27599 2018-01-08  Jeff Law  <law@redhat.com>
27601         PR rtl-optimizatin/81308
27602         * tree-switch-conversion.c (cfg_altered): New file scoped static.
27603         (process_switch): If group_case_labels makes a change, then set
27604         cfg_altered.
27605         (pass_convert_switch::execute): If a switch is converted, then
27606         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
27608         PR rtl-optimization/81308
27609         * recog.c (split_all_insns): Conditionally cleanup the CFG after
27610         splitting insns.
27612 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
27614         PR target/83663 - Revert r255946
27615         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
27616         generation for cases where splatting a value is not useful.
27617         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
27618         across a vec_duplicate and a paradoxical subreg forming a vector
27619         mode to a vec_concat.
27621 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27623         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
27624         -march=armv8.3-a variants.
27625         * config/arm/t-multilib: Likewise.
27626         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
27628 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27630         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
27631         to generate rtl.
27632         (cceq_ior_compare_complement): Give it a name so I can use it, and
27633         change boolean_or_operator predicate to boolean_operator so it can
27634         be used to generate a crand.
27635         (eqne): New code iterator.
27636         (bd/bd_neg): New code_attrs.
27637         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
27638         a single define_insn.
27639         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
27640         decrement (bdnzt/bdnzf/bdzt/bdzf).
27641         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
27642         with the new names of the branch decrement patterns, and added the
27643         names of the branch decrement conditional patterns.
27645 2018-01-08  Richard Biener  <rguenther@suse.de>
27647         PR tree-optimization/83563
27648         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
27649         cache.
27651 2018-01-08  Richard Biener  <rguenther@suse.de>
27653         PR middle-end/83713
27654         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
27656 2018-01-08  Richard Biener  <rguenther@suse.de>
27658         PR tree-optimization/83685
27659         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
27660         references to abnormals.
27662 2018-01-08  Richard Biener  <rguenther@suse.de>
27664         PR lto/83719
27665         * dwarf2out.c (output_indirect_strings): Handle empty
27666         skeleton_debug_str_hash.
27667         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
27669 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
27671         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
27672         (emit_store_direct): Likewise.
27673         (arc_trampoline_adjust_address): Likewise.
27674         (arc_asm_trampoline_template): New function.
27675         (arc_initialize_trampoline): Use asm_trampoline_template.
27676         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
27677         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
27678         * config/arc/arc.md (flush_icache): Delete pattern.
27680 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
27682         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
27683         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
27684         munaligned-access.
27686 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
27688         PR target/83681
27689         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
27690         by not USED_FOR_TARGET.
27691         (make_pass_resolve_sw_modes): Likewise.
27693 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
27695         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
27696         USED_FOR_TARGET.
27698 2018-01-08  Richard Biener  <rguenther@suse.de>
27700         PR middle-end/83580
27701         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
27703 2018-01-08  Richard Biener  <rguenther@suse.de>
27705         PR middle-end/83517
27706         * match.pd ((t * 2) / 2) -> t): Add missing :c.
27708 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
27710         PR middle-end/81897
27711         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
27712         basic blocks with a small number of successors.
27713         (convert_control_dep_chain_into_preds): Improve handling of
27714         forwarder blocks.
27715         (dump_predicates): Split apart into...
27716         (dump_pred_chain): ...here...
27717         (dump_pred_info): ...and here.
27718         (can_one_predicate_be_invalidated_p): Add debugging printfs.
27719         (can_chain_union_be_invalidated_p): Improve check for invalidation
27720         of paths.
27721         (uninit_uses_cannot_happen): Avoid unnecessary if
27722         convert_control_dep_chain_into_preds yielded nothing.
27724 2018-01-06  Martin Sebor  <msebor@redhat.com>
27726         PR tree-optimization/83640
27727         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
27728         subtracting negative offset from size.
27729         (builtin_access::overlap): Adjust offset bounds of the access to fall
27730         within the size of the object if possible.
27732 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
27734         PR rtl-optimization/83699
27735         * expmed.c (extract_bit_field_1): Restrict the vector usage of
27736         extract_bit_field_as_subreg to cases in which the extracted
27737         value is also a vector.
27739         * lra-constraints.c (process_alt_operands): Test for the equivalence
27740         substitutions when detecting a possible reload cycle.
27742 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
27744         PR debug/83480
27745         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
27746         by default if flag_selective_schedling{,2}.  Formatting fixes.
27748         PR rtl-optimization/83682
27749         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
27750         if it has non-VECTOR_MODE element mode.
27751         (vec_duplicate_p): Likewise.
27753         PR middle-end/83694
27754         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
27755         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
27757 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
27759         PR target/83604
27760         * config/i386/i386-builtin.def
27761         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
27762         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
27763         Require also OPTION_MASK_ISA_AVX512F in addition to
27764         OPTION_MASK_ISA_GFNI.
27765         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
27766         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
27767         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
27768         to OPTION_MASK_ISA_GFNI.
27769         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
27770         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
27771         OPTION_MASK_ISA_AVX512BW.
27772         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
27773         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
27774         addition to OPTION_MASK_ISA_GFNI.
27775         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
27776         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
27777         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
27778         to OPTION_MASK_ISA_GFNI.
27779         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
27780         a requirement for all ISAs rather than any of them with a few
27781         exceptions.
27782         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
27783         processing.
27784         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
27785         bitmasks to be enabled with 3 exceptions, instead of requiring any
27786         enabled ISA with lots of exceptions.
27787         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
27788         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
27789         Change avx512bw in isa attribute to avx512f.
27790         * config/i386/sgxintrin.h: Add license boilerplate.
27791         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
27792         to __AVX512F__ and __AVX512VL to __AVX512VL__.
27793         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
27794         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
27795         defined.
27796         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
27797         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
27798         temporarily sse2 rather than sse if not enabled already.
27800         PR target/83604
27801         * config/i386/sse.md (VI248_VLBW): Rename to ...
27802         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
27803         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
27804         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
27805         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
27806         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
27807         mode iterator instead of VI248_VLBW.
27809 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
27811         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
27812         (record_modified): Skip clobbers; add debug output.
27813         (param_change_prob): Use sreal frequencies.
27815 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
27817         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
27818         punt for user-aligned variables.
27820 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
27822         * tree-chrec.c (chrec_contains_symbols): Return true for
27823         POLY_INT_CST.
27825 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
27827         PR target/82439
27828         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
27829         of (x|y) == x for BICS pattern.
27831 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
27833         PR tree-optimization/83605
27834         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
27835         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
27836         can throw.
27838 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
27840         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
27841         * config/epiphany/rtems.h: New file.
27843 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27844             Uros Bizjak  <ubizjak@gmail.com>
27846         PR target/83554
27847         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
27848         QIreg_operand instead of register_operand predicate.
27849         * config/i386/i386.c (ix86_rop_should_change_byte_p,
27850         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
27851         comments instead of -fmitigate[-_]rop.
27853 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27855         PR bootstrap/81926
27856         * cgraphunit.c (symbol_table::compile): Switch to text_section
27857         before calling assembly_start debug hook.
27858         * run-rtl-passes.c (run_rtl_passes): Likewise.
27859         Include output.h.
27861 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27863         * tree-vrp.c (extract_range_from_binary_expr_1): Check
27864         range_int_cst_p rather than !symbolic_range_p before calling
27865         extract_range_from_multiplicative_op_1.
27867 2018-01-04  Jeff Law  <law@redhat.com>
27869         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
27870         redundant test in assertion.
27872 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27874         * doc/rtl.texi: Document machine_mode wrapper classes.
27876 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27878         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
27879         using tree_to_uhwi.
27881 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27883         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
27884         the VEC_PERM_EXPR fold to fail.
27886 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27888         PR debug/83585
27889         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
27890         to switched_sections.
27892 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27894         PR target/83680
27895         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
27896         test for d.testing.
27898 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
27900         PR target/83387
27901         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
27902         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
27904 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27906         PR debug/83666
27907         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
27908         is BLKmode and bitpos not zero or mode change is needed.
27910 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
27912         PR target/83675
27913         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
27914         TARGET_VIS2.
27916 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
27918         PR target/83628
27919         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
27920         instead of MULT rtx.  Update all corresponding splitters.
27921         (*saddl_se): Ditto.
27922         (*ssub<modesuffix>): Ditto.
27923         (*ssubl_se): Ditto.
27924         (*cmp_sadd_di): Update split patterns.
27925         (*cmp_sadd_si): Ditto.
27926         (*cmp_sadd_sidi): Ditto.
27927         (*cmp_ssub_di): Ditto.
27928         (*cmp_ssub_si): Ditto.
27929         (*cmp_ssub_sidi): Ditto.
27930         * config/alpha/predicates.md (const23_operand): New predicate.
27931         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
27932         Look for ASHIFT, not MULT inner operand.
27933         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
27935 2018-01-04  Martin Liska  <mliska@suse.cz>
27937         PR gcov-profile/83669
27938         * gcov.c (output_intermediate_file): Add version to intermediate
27939         gcov file.
27940         * doc/gcov.texi: Document new field 'version' in intermediate
27941         file format. Fix location of '-k' option of gcov command.
27943 2018-01-04  Martin Liska  <mliska@suse.cz>
27945         PR ipa/82352
27946         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
27948 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
27950         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
27952 2018-01-03  Martin Sebor  <msebor@redhat.com>
27954         PR tree-optimization/83655
27955         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
27956         checking calls with invalid arguments.
27958 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27960         * tree-vect-stmts.c (vect_get_store_rhs): New function.
27961         (vectorizable_mask_load_store): Delete.
27962         (vectorizable_call): Return false for masked loads and stores.
27963         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
27964         instead of gimple_assign_rhs1.
27965         (vectorizable_load): Handle IFN_MASK_LOAD.
27966         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
27968 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27970         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
27971         split out from..,
27972         (vectorizable_mask_load_store): ...here.
27973         (vectorizable_load): ...and here.
27975 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27977         * tree-vect-stmts.c (vect_build_all_ones_mask)
27978         (vect_build_zero_merge_argument): New functions, split out from...
27979         (vectorizable_load): ...here.
27981 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27983         * tree-vect-stmts.c (vect_check_store_rhs): New function,
27984         split out from...
27985         (vectorizable_mask_load_store): ...here.
27986         (vectorizable_store): ...and here.
27988 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27990         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
27991         split out from...
27992         (vectorizable_mask_load_store): ...here.
27994 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
27996         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
27997         (vect_model_store_cost): Take a vec_load_store_type instead of a
27998         vect_def_type.
27999         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
28000         (vect_model_store_cost): Take a vec_load_store_type instead of a
28001         vect_def_type.
28002         (vectorizable_mask_load_store): Update accordingly.
28003         (vectorizable_store): Likewise.
28004         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
28006 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28008         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
28009         IFN_MASK_LOAD calls here rather than...
28010         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
28012 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28013             Alan Hayward  <alan.hayward@arm.com>
28014             David Sherwood  <david.sherwood@arm.com>
28016         * expmed.c (extract_bit_field_1): For vector extracts,
28017         fall back to extract_bit_field_as_subreg if vec_extract
28018         isn't available.
28020 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28021             Alan Hayward  <alan.hayward@arm.com>
28022             David Sherwood  <david.sherwood@arm.com>
28024         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
28025         they are variable or constant sized.
28026         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
28027         slots for constant-sized data.
28029 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28030             Alan Hayward  <alan.hayward@arm.com>
28031             David Sherwood  <david.sherwood@arm.com>
28033         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
28034         handling COND_EXPRs with boolean comparisons, try to find a better
28035         basis for the mask type than the boolean itself.
28037 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28039         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
28040         is calculated and how it can be overridden.
28041         * genmodes.c (max_bitsize_mode_any_mode): New variable.
28042         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
28043         if defined.
28044         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
28045         if nonzero.
28047 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28048             Alan Hayward  <alan.hayward@arm.com>
28049             David Sherwood  <david.sherwood@arm.com>
28051         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
28052         Remove the mode argument.
28053         (aarch64_simd_valid_immediate): Remove the mode and inverse
28054         arguments.
28055         * config/aarch64/iterators.md (bitsize): New iterator.
28056         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
28057         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
28058         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
28059         aarch64_simd_valid_immediate.
28060         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
28061         (aarch64_reg_or_bic_imm): Likewise.
28062         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
28063         with an insn_type enum and msl with a modifier_type enum.
28064         Replace element_width with a scalar_mode.  Change the shift
28065         to unsigned int.  Add constructors for scalar_float_mode and
28066         scalar_int_mode elements.
28067         (aarch64_vect_float_const_representable_p): Delete.
28068         (aarch64_can_const_movi_rtx_p)
28069         (aarch64_simd_scalar_immediate_valid_for_move)
28070         (aarch64_simd_make_constant): Update call to
28071         aarch64_simd_valid_immediate.
28072         (aarch64_advsimd_valid_immediate_hs): New function.
28073         (aarch64_advsimd_valid_immediate): Likewise.
28074         (aarch64_simd_valid_immediate): Remove mode and inverse
28075         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
28076         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
28077         and aarch64_float_const_representable_p on the result.
28078         (aarch64_output_simd_mov_immediate): Remove mode argument.
28079         Update call to aarch64_simd_valid_immediate and use of
28080         simd_immediate_info.
28081         (aarch64_output_scalar_simd_mov_immediate): Update call
28082         accordingly.
28084 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28085             Alan Hayward  <alan.hayward@arm.com>
28086             David Sherwood  <david.sherwood@arm.com>
28088         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
28089         (mode_nunits): Likewise CONST_MODE_NUNITS.
28090         * machmode.def (ADJUST_NUNITS): Document.
28091         * genmodes.c (mode_data::need_nunits_adj): New field.
28092         (blank_mode): Update accordingly.
28093         (adj_nunits): New variable.
28094         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
28095         parameter.
28096         (emit_mode_size_inline): Set need_bytesize_adj for all modes
28097         listed in adj_nunits.
28098         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
28099         listed in adj_nunits.  Don't emit case statements for such modes.
28100         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
28101         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
28102         nothing if adj_nunits is nonnull.
28103         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
28104         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
28105         (emit_mode_fbit): Update use of print_maybe_const_decl.
28106         (emit_move_size): Likewise.  Treat the array as non-const
28107         if adj_nunits.
28108         (emit_mode_adjustments): Handle adj_nunits.
28110 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28112         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
28113         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
28114         (VECTOR_MODES): Use it.
28115         (make_vector_modes): Take the prefix as an argument.
28117 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28118             Alan Hayward  <alan.hayward@arm.com>
28119             David Sherwood  <david.sherwood@arm.com>
28121         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
28122         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
28123         for MODE_VECTOR_BOOL.
28124         * machmode.def (VECTOR_BOOL_MODE): Document.
28125         * genmodes.c (VECTOR_BOOL_MODE): New macro.
28126         (make_vector_bool_mode): New function.
28127         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
28128         MODE_VECTOR_BOOL.
28129         * lto-streamer-in.c (lto_input_mode_table): Likewise.
28130         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
28131         Likewise.
28132         * stor-layout.c (int_mode_for_mode): Likewise.
28133         * tree.c (build_vector_type_for_mode): Likewise.
28134         * varasm.c (output_constant_pool_2): Likewise.
28135         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
28136         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
28137         for MODE_VECTOR_BOOL.
28138         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
28139         of mode class checks.
28140         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
28141         instead of a list of mode class checks.
28142         (expand_vector_scalar_condition): Likewise.
28143         (type_for_widest_vector_mode): Handle BImode as an inner mode.
28145 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28146             Alan Hayward  <alan.hayward@arm.com>
28147             David Sherwood  <david.sherwood@arm.com>
28149         * machmode.h (mode_size): Change from unsigned short to
28150         poly_uint16_pod.
28151         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
28152         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28153         or if measurement_type is not polynomial.
28154         (fixed_size_mode::includes_p): Check for constant-sized modes.
28155         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
28156         return a poly_uint16 rather than an unsigned short.
28157         (emit_mode_size): Change the type of mode_size from unsigned short
28158         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
28159         (emit_mode_adjustments): Cope with polynomial vector sizes.
28160         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28161         for GET_MODE_SIZE.
28162         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28163         for GET_MODE_SIZE.
28164         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
28165         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
28166         * caller-save.c (setup_save_areas): Likewise.
28167         (replace_reg_with_saved_mem): Likewise.
28168         * calls.c (emit_library_call_value_1): Likewise.
28169         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
28170         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
28171         (gen_lowpart_for_combine): Likewise.
28172         * convert.c (convert_to_integer_1): Likewise.
28173         * cse.c (equiv_constant, cse_insn): Likewise.
28174         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
28175         (cselib_subst_to_values): Likewise.
28176         * dce.c (word_dce_process_block): Likewise.
28177         * df-problems.c (df_word_lr_mark_ref): Likewise.
28178         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
28179         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
28180         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
28181         (rtl_for_decl_location): Likewise.
28182         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
28183         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
28184         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
28185         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
28186         (expand_expr_real_1): Likewise.
28187         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
28188         (pad_below): Likewise.
28189         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
28190         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
28191         * ira.c (get_subreg_tracking_sizes): Likewise.
28192         * ira-build.c (ira_create_allocno_objects): Likewise.
28193         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
28194         (ira_sort_regnos_for_alter_reg): Likewise.
28195         * ira-costs.c (record_operand_costs): Likewise.
28196         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
28197         (resolve_simple_move): Likewise.
28198         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
28199         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
28200         (lra_constraints): Likewise.
28201         (CONST_POOL_OK_P): Reject variable-sized modes.
28202         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
28203         (add_pseudo_to_slot, lra_spill): Likewise.
28204         * omp-low.c (omp_clause_aligned_alignment): Likewise.
28205         * optabs-query.c (get_best_extraction_insn): Likewise.
28206         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
28207         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
28208         (expand_mult_highpart, valid_multiword_target_p): Likewise.
28209         * recog.c (offsettable_address_addr_space_p): Likewise.
28210         * regcprop.c (maybe_mode_change): Likewise.
28211         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
28212         * regrename.c (build_def_use): Likewise.
28213         * regstat.c (dump_reg_info): Likewise.
28214         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
28215         (find_reloads, find_reloads_subreg_address): Likewise.
28216         * reload1.c (eliminate_regs_1): Likewise.
28217         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
28218         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
28219         (simplify_binary_operation_1, simplify_subreg): Likewise.
28220         * targhooks.c (default_function_arg_padding): Likewise.
28221         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
28222         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
28223         (verify_gimple_assign_ternary): Likewise.
28224         * tree-inline.c (estimate_move_cost): Likewise.
28225         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28226         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
28227         (get_address_cost_ainc): Likewise.
28228         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
28229         (vect_supportable_dr_alignment): Likewise.
28230         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
28231         (vectorizable_reduction): Likewise.
28232         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
28233         (vectorizable_operation, vectorizable_load): Likewise.
28234         * tree.c (build_same_sized_truth_vector_type): Likewise.
28235         * valtrack.c (cleanup_auto_inc_dec): Likewise.
28236         * var-tracking.c (emit_note_insn_var_location): Likewise.
28237         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
28238         (ADDR_VEC_ALIGN): Likewise.
28240 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28241             Alan Hayward  <alan.hayward@arm.com>
28242             David Sherwood  <david.sherwood@arm.com>
28244         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
28245         unsigned short.
28246         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
28247         or if measurement_type is polynomial.
28248         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
28249         * combine.c (make_extraction): Likewise.
28250         * dse.c (find_shift_sequence): Likewise.
28251         * dwarf2out.c (mem_loc_descriptor): Likewise.
28252         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
28253         (extract_bit_field, extract_low_bits): Likewise.
28254         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
28255         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
28256         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
28257         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
28258         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
28259         * reload.c (find_reloads): Likewise.
28260         * reload1.c (alter_reg): Likewise.
28261         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
28262         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
28263         * tree-if-conv.c (predicate_mem_writes): Likewise.
28264         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
28265         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
28266         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
28267         * valtrack.c (dead_debug_insert_temp): Likewise.
28268         * varasm.c (mergeable_constant_section): Likewise.
28269         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
28271 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28272             Alan Hayward  <alan.hayward@arm.com>
28273             David Sherwood  <david.sherwood@arm.com>
28275         * expr.c (expand_assignment): Cope with polynomial mode sizes
28276         when assigning to a CONCAT.
28278 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28279             Alan Hayward  <alan.hayward@arm.com>
28280             David Sherwood  <david.sherwood@arm.com>
28282         * machmode.h (mode_precision): Change from unsigned short to
28283         poly_uint16_pod.
28284         (mode_to_precision): Return a poly_uint16 rather than an unsigned
28285         short.
28286         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
28287         or if measurement_type is not polynomial.
28288         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
28289         in which the mode is already known to be a scalar_int_mode.
28290         * genmodes.c (emit_mode_precision): Change the type of mode_precision
28291         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
28292         initializer.
28293         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28294         for GET_MODE_PRECISION.
28295         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28296         for GET_MODE_PRECISION.
28297         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
28298         as polynomial.
28299         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
28300         (expand_field_assignment, make_extraction): Likewise.
28301         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
28302         (get_last_value): Likewise.
28303         * convert.c (convert_to_integer_1): Likewise.
28304         * cse.c (cse_insn): Likewise.
28305         * expr.c (expand_expr_real_1): Likewise.
28306         * lra-constraints.c (simplify_operand_subreg): Likewise.
28307         * optabs-query.c (can_atomic_load_p): Likewise.
28308         * optabs.c (expand_atomic_load): Likewise.
28309         (expand_atomic_store): Likewise.
28310         * ree.c (combine_reaching_defs): Likewise.
28311         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
28312         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
28313         * tree.h (type_has_mode_precision_p): Likewise.
28314         * ubsan.c (instrument_si_overflow): Likewise.
28316 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28317             Alan Hayward  <alan.hayward@arm.com>
28318             David Sherwood  <david.sherwood@arm.com>
28320         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
28321         polynomial numbers of units.
28322         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
28323         (valid_vector_subparts_p): New function.
28324         (build_vector_type): Remove temporary shim and take the number
28325         of units as a poly_uint64 rather than an int.
28326         (build_opaque_vector_type): Take the number of units as a
28327         poly_uint64 rather than an int.
28328         * tree.c (build_vector_from_ctor): Handle polynomial
28329         TYPE_VECTOR_SUBPARTS.
28330         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
28331         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
28332         (build_vector_from_val): If the number of units is variable,
28333         use build_vec_duplicate_cst for constant operands and
28334         VEC_DUPLICATE_EXPR otherwise.
28335         (make_vector_type): Remove temporary is_constant ().
28336         (build_vector_type, build_opaque_vector_type): Take the number of
28337         units as a poly_uint64 rather than an int.
28338         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
28339         VECTOR_CST_NELTS.
28340         * cfgexpand.c (expand_debug_expr): Likewise.
28341         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
28342         (store_constructor, expand_expr_real_1): Likewise.
28343         (const_scalar_mask_from_tree): Likewise.
28344         * fold-const-call.c (fold_const_reduction): Likewise.
28345         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
28346         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
28347         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
28348         (fold_relational_const): Likewise.
28349         (native_interpret_vector): Likewise.  Change the size from an
28350         int to an unsigned int.
28351         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
28352         TYPE_VECTOR_SUBPARTS.
28353         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
28354         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
28355         duplicating a non-constant operand into a variable-length vector.
28356         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
28357         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
28358         * ipa-icf.c (sem_variable::equals): Likewise.
28359         * match.pd: Likewise.
28360         * omp-simd-clone.c (simd_clone_subparts): Likewise.
28361         * print-tree.c (print_node): Likewise.
28362         * stor-layout.c (layout_type): Likewise.
28363         * targhooks.c (default_builtin_vectorization_cost): Likewise.
28364         * tree-cfg.c (verify_gimple_comparison): Likewise.
28365         (verify_gimple_assign_binary): Likewise.
28366         (verify_gimple_assign_ternary): Likewise.
28367         (verify_gimple_assign_single): Likewise.
28368         * tree-pretty-print.c (dump_generic_node): Likewise.
28369         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28370         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
28371         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
28372         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
28373         (vect_shift_permute_load_chain): Likewise.
28374         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
28375         (expand_vector_condition, optimize_vector_constructor): Likewise.
28376         (lower_vec_perm, get_compute_type): Likewise.
28377         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
28378         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
28379         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
28380         (vect_recog_mask_conversion_pattern): Likewise.
28381         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
28382         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
28383         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28384         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
28385         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
28386         (vectorizable_shift, vectorizable_operation, vectorizable_store)
28387         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
28388         (supportable_widening_operation): Likewise.
28389         (supportable_narrowing_operation): Likewise.
28390         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
28391         Likewise.
28392         * varasm.c (output_constant): Likewise.
28394 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28395             Alan Hayward  <alan.hayward@arm.com>
28396             David Sherwood  <david.sherwood@arm.com>
28398         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
28399         so that both the length == 3 and length != 3 cases set up their
28400         own permute vectors.  Add comments explaining why we know the
28401         number of elements is constant.
28402         (vect_permute_load_chain): Likewise.
28404 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28405             Alan Hayward  <alan.hayward@arm.com>
28406             David Sherwood  <david.sherwood@arm.com>
28408         * machmode.h (mode_nunits): Change from unsigned char to
28409         poly_uint16_pod.
28410         (ONLY_FIXED_SIZE_MODES): New macro.
28411         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
28412         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
28413         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
28414         New typedefs.
28415         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
28416         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
28417         or if measurement_type is not polynomial.
28418         * genmodes.c (ZERO_COEFFS): New macro.
28419         (emit_mode_nunits_inline): Make mode_nunits_inline return a
28420         poly_uint16.
28421         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
28422         Use ZERO_COEFFS when emitting initializers.
28423         * data-streamer.h (bp_pack_poly_value): New function.
28424         (bp_unpack_poly_value): Likewise.
28425         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
28426         for GET_MODE_NUNITS.
28427         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
28428         for GET_MODE_NUNITS.
28429         * tree.c (make_vector_type): Remove temporary shim and make
28430         the real function take the number of units as a poly_uint64
28431         rather than an int.
28432         (build_vector_type_for_mode): Handle polynomial nunits.
28433         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
28434         * emit-rtl.c (const_vec_series_p_1): Likewise.
28435         (gen_rtx_CONST_VECTOR): Likewise.
28436         * fold-const.c (test_vec_duplicate_folding): Likewise.
28437         * genrecog.c (validate_pattern): Likewise.
28438         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
28439         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
28440         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
28441         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
28442         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
28443         * rtlanal.c (subreg_get_info): Likewise.
28444         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28445         (vect_grouped_load_supported): Likewise.
28446         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
28447         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
28448         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
28449         (simplify_const_unary_operation, simplify_binary_operation_1)
28450         (simplify_const_binary_operation, simplify_ternary_operation)
28451         (test_vector_ops_duplicate, test_vector_ops): Likewise.
28452         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
28453         instead of CONST_VECTOR_NUNITS.
28454         * varasm.c (output_constant_pool_2): Likewise.
28455         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
28456         explicit-encoded elements in the XVEC for variable-length vectors.
28458 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28460         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
28462 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28463             Alan Hayward  <alan.hayward@arm.com>
28464             David Sherwood  <david.sherwood@arm.com>
28466         * coretypes.h (fixed_size_mode): Declare.
28467         (fixed_size_mode_pod): New typedef.
28468         * builtins.h (target_builtins::x_apply_args_mode)
28469         (target_builtins::x_apply_result_mode): Change type to
28470         fixed_size_mode_pod.
28471         * builtins.c (apply_args_size, apply_result_size, result_vector)
28472         (expand_builtin_apply_args_1, expand_builtin_apply)
28473         (expand_builtin_return): Update accordingly.
28475 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28477         * cse.c (hash_rtx_cb): Hash only the encoded elements.
28478         * cselib.c (cselib_hash_rtx): Likewise.
28479         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
28480         CONST_VECTOR encoding.
28482 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28483             Jeff Law  <law@redhat.com>
28485         PR target/83641
28486         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
28487         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
28488         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
28489         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
28491         PR target/83641
28492         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
28493         explicitly probe *sp in a noreturn function if there were any callee
28494         register saves or frame pointer is needed.
28496 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28498         PR debug/83621
28499         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
28500         BLKmode for ternary, binary or unary expressions.
28502         PR debug/83645
28503         * var-tracking.c (delete_vta_debug_insn): New inline function.
28504         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
28505         insns from get_insns () to NULL instead of each bb separately.
28506         Use delete_vta_debug_insn.  No longer static.
28507         (vt_debug_insns_local, variable_tracking_main_1): Adjust
28508         delete_vta_debug_insns callers.
28509         * rtl.h (delete_vta_debug_insns): Declare.
28510         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
28511         instead of variable_tracking_main.
28513 2018-01-03  Martin Sebor  <msebor@redhat.com>
28515         PR tree-optimization/83603
28516         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
28517         arguments past the endof the argument list in functions declared
28518         without a prototype.
28519         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
28520         Avoid checking when arguments are null.
28522 2018-01-03  Martin Sebor  <msebor@redhat.com>
28524         PR c/83559
28525         * doc/extend.texi (attribute const): Fix a typo.
28526         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
28527         issuing -Wsuggest-attribute for void functions.
28529 2018-01-03  Martin Sebor  <msebor@redhat.com>
28531         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
28532         offset_int::from instead of wide_int::to_shwi.
28533         (maybe_diag_overlap): Remove assertion.
28534         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
28535         * gimple-ssa-sprintf.c (format_directive): Same.
28536         (parse_directive): Same.
28537         (sprintf_dom_walker::compute_format_length): Same.
28538         (try_substitute_return_value): Same.
28540 2018-01-03  Jeff Law  <law@redhat.com>
28542         PR middle-end/83654
28543         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
28544         non-constant residual for zero at runtime and avoid probing in
28545         that case.  Reorganize code for trailing problem to mirror handling
28546         of the residual.
28548 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
28550         PR tree-optimization/83501
28551         * tree-ssa-strlen.c (get_string_cst): New.
28552         (handle_char_store): Call get_string_cst.
28554 2018-01-03  Martin Liska  <mliska@suse.cz>
28556         PR tree-optimization/83593
28557         * tree-ssa-strlen.c: Include tree-cfg.h.
28558         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
28559         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
28560         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
28561         to false.
28562         (strlen_dom_walker::before_dom_children): Call
28563         gimple_purge_dead_eh_edges. Dump tranformation with details
28564         dump flags.
28565         (strlen_dom_walker::before_dom_children): Update call by adding
28566         new argument cleanup_eh.
28567         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
28569 2018-01-03  Martin Liska  <mliska@suse.cz>
28571         PR ipa/83549
28572         * cif-code.def (VARIADIC_THUNK): New enum value.
28573         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
28574         thunks.
28576 2018-01-03  Jan Beulich  <jbeulich@suse.com>
28578         * sse.md (mov<mode>_internal): Tighten condition for when to use
28579         vmovdqu<ssescalarsize> for TI and OI modes.
28581 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28583         Update copyright years.
28585 2018-01-03  Martin Liska  <mliska@suse.cz>
28587         PR ipa/83594
28588         * ipa-visibility.c (function_and_variable_visibility): Skip
28589         functions with noipa attribure.
28591 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
28593         * gcc.c (process_command): Update copyright notice dates.
28594         * gcov-dump.c (print_version): Ditto.
28595         * gcov.c (print_version): Ditto.
28596         * gcov-tool.c (print_version): Ditto.
28597         * gengtype.c (create_file): Ditto.
28598         * doc/cpp.texi: Bump @copying's copyright year.
28599         * doc/cppinternals.texi: Ditto.
28600         * doc/gcc.texi: Ditto.
28601         * doc/gccint.texi: Ditto.
28602         * doc/gcov.texi: Ditto.
28603         * doc/install.texi: Ditto.
28604         * doc/invoke.texi: Ditto.
28606 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28608         * vector-builder.h (vector_builder::m_full_nelts): Change from
28609         unsigned int to poly_uint64.
28610         (vector_builder::full_nelts): Update prototype accordingly.
28611         (vector_builder::new_vector): Likewise.
28612         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
28613         (vector_builder::operator ==): Likewise.
28614         (vector_builder::finalize): Likewise.
28615         * int-vector-builder.h (int_vector_builder::int_vector_builder):
28616         Take the number of elements as a poly_uint64 rather than an
28617         unsigned int.
28618         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
28619         from unsigned int to poly_uint64.
28620         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
28621         (vec_perm_indices::new_vector): Likewise.
28622         (vec_perm_indices::length): Likewise.
28623         (vec_perm_indices::nelts_per_input): Likewise.
28624         (vec_perm_indices::input_nelts): Likewise.
28625         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
28626         number of elements per input as a poly_uint64 rather than an
28627         unsigned int.  Use the original encoding for variable-length
28628         vectors, rather than clamping each individual element.
28629         For the second and subsequent elements in each pattern,
28630         clamp the step and base before clamping their sum.
28631         (vec_perm_indices::series_p): Handle polynomial element counts.
28632         (vec_perm_indices::all_in_range_p): Likewise.
28633         (vec_perm_indices_to_tree): Likewise.
28634         (vec_perm_indices_to_rtx): Likewise.
28635         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
28636         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
28637         (tree_vector_builder::new_binary_operation): Handle polynomial
28638         element counts.  Return false if we need to know the number
28639         of elements at compile time.
28640         * fold-const.c (fold_vec_perm): Punt if the number of elements
28641         isn't known at compile time.
28643 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28645         * vec-perm-indices.h (vec_perm_builder): Change element type
28646         from HOST_WIDE_INT to poly_int64.
28647         (vec_perm_indices::element_type): Update accordingly.
28648         (vec_perm_indices::clamp): Handle polynomial element_types.
28649         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
28650         (vec_perm_indices::all_in_range_p): Likewise.
28651         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
28652         than shwi trees.
28653         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
28654         polynomial vec_perm_indices element types.
28655         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
28656         * fold-const.c (fold_vec_perm): Likewise.
28657         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
28658         * tree-vect-generic.c (lower_vec_perm): Likewise.
28659         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
28660         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
28661         element type to HOST_WIDE_INT.
28663 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28664             Alan Hayward  <alan.hayward@arm.com>
28665             David Sherwood  <david.sherwood@arm.com>
28667         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
28668         rather than an int.  Use plus_constant.
28669         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
28670         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
28672 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28673             Alan Hayward  <alan.hayward@arm.com>
28674             David Sherwood  <david.sherwood@arm.com>
28676         * calls.c (emit_call_1, expand_call): Change struct_value_size from
28677         a HOST_WIDE_INT to a poly_int64.
28679 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28680             Alan Hayward  <alan.hayward@arm.com>
28681             David Sherwood  <david.sherwood@arm.com>
28683         * calls.c (load_register_parameters): Cope with polynomial
28684         mode sizes.  Require a constant size for BLKmode parameters
28685         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
28686         forces a parameter to be padded at the lsb end in order to
28687         fill a complete number of words, require the parameter size
28688         to be ordered wrt UNITS_PER_WORD.
28690 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28691             Alan Hayward  <alan.hayward@arm.com>
28692             David Sherwood  <david.sherwood@arm.com>
28694         * reload1.c (spill_stack_slot_width): Change element type
28695         from unsigned int to poly_uint64_pod.
28696         (alter_reg): Treat mode sizes as polynomial.
28698 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28699             Alan Hayward  <alan.hayward@arm.com>
28700             David Sherwood  <david.sherwood@arm.com>
28702         * reload.c (complex_word_subreg_p): New function.
28703         (reload_inner_reg_of_subreg, push_reload): Use it.
28705 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28706             Alan Hayward  <alan.hayward@arm.com>
28707             David Sherwood  <david.sherwood@arm.com>
28709         * lra-constraints.c (process_alt_operands): Reject matched
28710         operands whose sizes aren't ordered.
28711         (match_reload): Refer to this check here.
28713 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28714             Alan Hayward  <alan.hayward@arm.com>
28715             David Sherwood  <david.sherwood@arm.com>
28717         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
28718         that the mode size is in the set {1, 2, 4, 8, 16}.
28720 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28721             Alan Hayward  <alan.hayward@arm.com>
28722             David Sherwood  <david.sherwood@arm.com>
28724         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
28725         Use plus_constant instead of gen_rtx_PLUS.
28727 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28728             Alan Hayward  <alan.hayward@arm.com>
28729             David Sherwood  <david.sherwood@arm.com>
28731         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
28732         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
28733         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
28734         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
28735         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
28736         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
28737         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
28738         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
28739         * config/i386/i386.c (ix86_push_rounding): ...this new function.
28740         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
28741         a poly_int64.
28742         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
28743         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
28744         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
28745         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
28746         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
28747         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
28748         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
28749         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
28750         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
28751         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
28752         function.
28753         * expr.c (emit_move_resolve_push): Treat the input and result
28754         of PUSH_ROUNDING as a poly_int64.
28755         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
28756         (emit_push_insn): Likewise.
28757         * lra-eliminations.c (mark_not_eliminable): Likewise.
28758         * recog.c (push_operand): Likewise.
28759         * reload1.c (elimination_effects): Likewise.
28760         * rtlanal.c (nonzero_bits1): Likewise.
28761         * calls.c (store_one_arg): Likewise.  Require the padding to be
28762         known at compile time.
28764 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28765             Alan Hayward  <alan.hayward@arm.com>
28766             David Sherwood  <david.sherwood@arm.com>
28768         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
28769         Use plus_constant instead of gen_rtx_PLUS.
28771 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28772             Alan Hayward  <alan.hayward@arm.com>
28773             David Sherwood  <david.sherwood@arm.com>
28775         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
28776         rather than an int.
28778 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28779             Alan Hayward  <alan.hayward@arm.com>
28780             David Sherwood  <david.sherwood@arm.com>
28782         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
28783         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
28784         via stack temporaries.  Treat the mode size as polynomial too.
28786 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28787             Alan Hayward  <alan.hayward@arm.com>
28788             David Sherwood  <david.sherwood@arm.com>
28790         * expr.c (expand_expr_real_2): When handling conversions involving
28791         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
28792         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
28793         as a poly_uint64 too.
28795 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28796             Alan Hayward  <alan.hayward@arm.com>
28797             David Sherwood  <david.sherwood@arm.com>
28799         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
28801 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28802             Alan Hayward  <alan.hayward@arm.com>
28803             David Sherwood  <david.sherwood@arm.com>
28805         * combine.c (can_change_dest_mode): Handle polynomial
28806         REGMODE_NATURAL_SIZE.
28807         * expmed.c (store_bit_field_1): Likewise.
28808         * expr.c (store_constructor): Likewise.
28809         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
28810         and polynomial REGMODE_NATURAL_SIZE.
28811         (gen_lowpart_common): Likewise.
28812         * reginfo.c (record_subregs_of_mode): Likewise.
28813         * rtlanal.c (read_modify_subreg_p): Likewise.
28815 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28816             Alan Hayward  <alan.hayward@arm.com>
28817             David Sherwood  <david.sherwood@arm.com>
28819         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
28820         numbers of elements.
28822 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28823             Alan Hayward  <alan.hayward@arm.com>
28824             David Sherwood  <david.sherwood@arm.com>
28826         * match.pd: Cope with polynomial numbers of vector elements.
28828 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28829             Alan Hayward  <alan.hayward@arm.com>
28830             David Sherwood  <david.sherwood@arm.com>
28832         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
28833         in a POINTER_PLUS_EXPR.
28835 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28836             Alan Hayward  <alan.hayward@arm.com>
28837             David Sherwood  <david.sherwood@arm.com>
28839         * omp-simd-clone.c (simd_clone_subparts): New function.
28840         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
28841         (ipa_simd_modify_function_body): Likewise.
28843 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28844             Alan Hayward  <alan.hayward@arm.com>
28845             David Sherwood  <david.sherwood@arm.com>
28847         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
28848         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
28849         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
28850         (expand_vector_condition, vector_element): Likewise.
28851         (subparts_gt): New function.
28852         (get_compute_type): Use subparts_gt.
28853         (count_type_subparts): Delete.
28854         (expand_vector_operations_1): Use subparts_gt instead of
28855         count_type_subparts.
28857 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28858             Alan Hayward  <alan.hayward@arm.com>
28859             David Sherwood  <david.sherwood@arm.com>
28861         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
28862         (vect_compile_time_alias): ...this new function.  Do the calculation
28863         on poly_ints rather than trees.
28864         (vect_prune_runtime_alias_test_list): Update call accordingly.
28866 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28867             Alan Hayward  <alan.hayward@arm.com>
28868             David Sherwood  <david.sherwood@arm.com>
28870         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
28871         numbers of units.
28872         (vect_schedule_slp_instance): Likewise.
28874 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28875             Alan Hayward  <alan.hayward@arm.com>
28876             David Sherwood  <david.sherwood@arm.com>
28878         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
28879         constant and extern definitions for variable-length vectors.
28880         (vect_get_constant_vectors): Note that the number of units
28881         is known to be constant.
28883 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28884             Alan Hayward  <alan.hayward@arm.com>
28885             David Sherwood  <david.sherwood@arm.com>
28887         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
28888         of units as polynomial.  Choose between WIDE and NARROW based
28889         on multiple_p.
28891 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28892             Alan Hayward  <alan.hayward@arm.com>
28893             David Sherwood  <david.sherwood@arm.com>
28895         * tree-vect-stmts.c (simd_clone_subparts): New function.
28896         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
28898 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28899             Alan Hayward  <alan.hayward@arm.com>
28900             David Sherwood  <david.sherwood@arm.com>
28902         * tree-vect-stmts.c (vectorizable_call): Treat the number of
28903         vectors as polynomial.  Use build_index_vector for
28904         IFN_GOMP_SIMD_LANE.
28906 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28907             Alan Hayward  <alan.hayward@arm.com>
28908             David Sherwood  <david.sherwood@arm.com>
28910         * tree-vect-stmts.c (get_load_store_type): Treat the number of
28911         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
28912         for variable-length vectors.
28913         (vectorizable_mask_load_store): Treat the number of units as
28914         polynomial, asserting that it is constant if the condition has
28915         already been enforced.
28916         (vectorizable_store, vectorizable_load): Likewise.
28918 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28919             Alan Hayward  <alan.hayward@arm.com>
28920             David Sherwood  <david.sherwood@arm.com>
28922         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
28923         of units as polynomial.  Punt if we can't tell at compile time
28924         which vector contains the final result.
28926 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28927             Alan Hayward  <alan.hayward@arm.com>
28928             David Sherwood  <david.sherwood@arm.com>
28930         * tree-vect-loop.c (vectorizable_induction): Treat the number
28931         of units as polynomial.  Punt on SLP inductions.  Use an integer
28932         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
28933         cast of such a series for variable-length floating-point
28934         reductions.
28936 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28937             Alan Hayward  <alan.hayward@arm.com>
28938             David Sherwood  <david.sherwood@arm.com>
28940         * tree.h (build_index_vector): Declare.
28941         * tree.c (build_index_vector): New function.
28942         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
28943         of units as polynomial, forcibly converting it to a constant if
28944         vectorizable_reduction has already enforced the condition.
28945         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
28946         to create a {1,2,3,...} vector.
28947         (vectorizable_reduction): Treat the number of units as polynomial.
28948         Choose vectype_in based on the largest scalar element size rather
28949         than the smallest number of units.  Enforce the restrictions
28950         relied on above.
28952 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28953             Alan Hayward  <alan.hayward@arm.com>
28954             David Sherwood  <david.sherwood@arm.com>
28956         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
28957         number of units as polynomial.
28959 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28960             Alan Hayward  <alan.hayward@arm.com>
28961             David Sherwood  <david.sherwood@arm.com>
28963         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
28964         * target.def (autovectorize_vector_sizes): Return the vector sizes
28965         by pointer, using vector_sizes rather than a bitmask.
28966         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
28967         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
28968         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
28969         Likewise.
28970         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
28971         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
28972         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
28973         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
28974         * omp-general.c (omp_max_vf): Likewise.
28975         * omp-low.c (omp_clause_aligned_alignment): Likewise.
28976         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
28977         * tree-vect-loop.c (vect_analyze_loop): Likewise.
28978         * tree-vect-slp.c (vect_slp_bb): Likewise.
28979         * doc/tm.texi: Regenerate.
28980         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
28981         to a poly_uint64.
28982         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
28983         the vector size as a poly_uint64 rather than an unsigned int.
28984         (current_vector_size): Change from an unsigned int to a poly_uint64.
28985         (get_vectype_for_scalar_type): Update accordingly.
28986         * tree.h (build_truth_vector_type): Take the size and number of
28987         units as a poly_uint64 rather than an unsigned int.
28988         (build_vector_type): Add a temporary overload that takes
28989         the number of units as a poly_uint64 rather than an unsigned int.
28990         * tree.c (make_vector_type): Likewise.
28991         (build_truth_vector_type): Take the number of units as a poly_uint64
28992         rather than an unsigned int.
28994 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
28995             Alan Hayward  <alan.hayward@arm.com>
28996             David Sherwood  <david.sherwood@arm.com>
28998         * target.def (get_mask_mode): Take the number of units and length
28999         as poly_uint64s rather than unsigned ints.
29000         * targhooks.h (default_get_mask_mode): Update accordingly.
29001         * targhooks.c (default_get_mask_mode): Likewise.
29002         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
29003         * doc/tm.texi: Regenerate.
29005 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29006             Alan Hayward  <alan.hayward@arm.com>
29007             David Sherwood  <david.sherwood@arm.com>
29009         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
29010         * omp-general.c (omp_max_vf): Likewise.
29011         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
29012         (expand_omp_simd): Handle polynomial safelen.
29013         * omp-low.c (omplow_simd_context): Add a default constructor.
29014         (omplow_simd_context::max_vf): Change from int to poly_uint64.
29015         (lower_rec_simd_input_clauses): Update accordingly.
29016         (lower_rec_input_clauses): Likewise.
29018 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29019             Alan Hayward  <alan.hayward@arm.com>
29020             David Sherwood  <david.sherwood@arm.com>
29022         * tree-vectorizer.h (vect_nunits_for_cost): New function.
29023         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
29024         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
29025         (vect_analyze_slp_cost): Likewise.
29026         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
29027         (vect_model_load_cost): Likewise.
29029 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29030             Alan Hayward  <alan.hayward@arm.com>
29031             David Sherwood  <david.sherwood@arm.com>
29033         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
29034         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
29035         from an unsigned int * to a poly_uint64_pod *.
29036         (calculate_unrolling_factor): New function.
29037         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
29039 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29040             Alan Hayward  <alan.hayward@arm.com>
29041             David Sherwood  <david.sherwood@arm.com>
29043         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
29044         from an unsigned int to a poly_uint64.
29045         (_loop_vec_info::slp_unrolling_factor): Likewise.
29046         (_loop_vec_info::vectorization_factor): Change from an int
29047         to a poly_uint64.
29048         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
29049         (vect_get_num_vectors): New function.
29050         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
29051         (vect_get_num_copies): Use vect_get_num_vectors.
29052         (vect_analyze_data_ref_dependences): Change max_vf from an int *
29053         to an unsigned int *.
29054         (vect_analyze_data_refs): Change min_vf from an int * to a
29055         poly_uint64 *.
29056         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29057         than an unsigned HOST_WIDE_INT.
29058         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
29059         (vect_analyze_data_ref_dependence): Change max_vf from an int *
29060         to an unsigned int *.
29061         (vect_analyze_data_ref_dependences): Likewise.
29062         (vect_compute_data_ref_alignment): Handle polynomial vf.
29063         (vect_enhance_data_refs_alignment): Likewise.
29064         (vect_prune_runtime_alias_test_list): Likewise.
29065         (vect_shift_permute_load_chain): Likewise.
29066         (vect_supportable_dr_alignment): Likewise.
29067         (dependence_distance_ge_vf): Take the vectorization factor as a
29068         poly_uint64 rather than an unsigned HOST_WIDE_INT.
29069         (vect_analyze_data_refs): Change min_vf from an int * to a
29070         poly_uint64 *.
29071         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
29072         vfm1 as a poly_uint64 rather than an int.  Make the same change
29073         for the returned bound_scalar.
29074         (vect_gen_vector_loop_niters): Handle polynomial vf.
29075         (vect_do_peeling): Likewise.  Update call to
29076         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
29077         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
29078         be constant.
29079         * tree-vect-loop.c (vect_determine_vectorization_factor)
29080         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
29081         (vect_get_known_peeling_cost): Likewise.
29082         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
29083         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
29084         (vect_transform_loop): Likewise.  Use the lowest possible VF when
29085         updating the upper bounds of the loop.
29086         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
29087         rather than an int.
29088         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
29089         polynomial unroll factors.
29090         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
29091         (vect_make_slp_decision): Likewise.
29092         (vect_supported_load_permutation_p): Likewise, and polynomial
29093         vf too.
29094         (vect_analyze_slp_cost): Handle polynomial vf.
29095         (vect_slp_analyze_node_operations): Likewise.
29096         (vect_slp_analyze_bb_1): Likewise.
29097         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
29098         than an unsigned HOST_WIDE_INT.
29099         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
29100         (vectorizable_load): Handle polynomial vf.
29101         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
29102         a poly_uint64.
29103         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
29105 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29106             Alan Hayward  <alan.hayward@arm.com>
29107             David Sherwood  <david.sherwood@arm.com>
29109         * match.pd: Handle bit operations involving three constants
29110         and try to fold one pair.
29112 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29114         * tree-vect-loop-manip.c: Include gimple-fold.h.
29115         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
29116         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
29117         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
29118         Add a path that uses a step of VF instead of 1, but disable it
29119         for now.
29120         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
29121         and niters_no_overflow parameters.  Update calls to
29122         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
29123         Create a new SSA name if the latter choses to use a ste other
29124         than zero, and return it via niters_vector_mult_vf_var.
29125         * tree-vect-loop.c (vect_transform_loop): Update calls to
29126         vect_do_peeling, vect_gen_vector_loop_niters and
29127         slpeel_make_loop_iterate_ntimes.
29128         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
29129         (vect_gen_vector_loop_niters): Update declarations after above changes.
29131 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
29133         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
29134         128-bit round to integer instructions.
29135         (ceil<mode>2): Likewise.
29136         (btrunc<mode>2): Likewise.
29137         (round<mode>2): Likewise.
29139 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29141         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
29142         unaligned VSX load/store on P8/P9.
29143         (expand_block_clear): Allow the use of unaligned VSX
29144         load/store on P8/P9.
29146 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29148         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
29149         New function.
29150         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
29151         swap associated with both a load and a store.
29153 2018-01-02  Andrew Waterman  <andrew@sifive.com>
29155         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
29156         * config/riscv/riscv.md (clear_cache): Use it.
29158 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
29160         * web.c: Remove out-of-date comment.
29162 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29164         * expr.c (fixup_args_size_notes): Check that any existing
29165         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
29166         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
29167         (emit_single_push_insn): ...here.
29169 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29171         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
29172         (const_vector_encoded_nelts): New function.
29173         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
29174         (const_vector_int_elt, const_vector_elt): Declare.
29175         * emit-rtl.c (const_vector_int_elt_1): New function.
29176         (const_vector_elt): Likewise.
29177         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
29178         of CONST_VECTOR_ELT.
29180 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29182         * expr.c: Include rtx-vector-builder.h.
29183         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
29184         directly on the tree encoding.
29185         (const_vector_from_tree): Likewise.
29186         * optabs.c: Include rtx-vector-builder.h.
29187         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
29188         sequence of "u" values.
29189         * vec-perm-indices.c: Include rtx-vector-builder.h.
29190         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
29191         directly on the vec_perm_indices encoding.
29193 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29195         * doc/rtl.texi (const_vector): Describe new encoding scheme.
29196         * Makefile.in (OBJS): Add rtx-vector-builder.o.
29197         * rtx-vector-builder.h: New file.
29198         * rtx-vector-builder.c: Likewise.
29199         * rtl.h (rtx_def::u2): Add a const_vector field.
29200         (CONST_VECTOR_NPATTERNS): New macro.
29201         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
29202         (CONST_VECTOR_DUPLICATE_P): Likewise.
29203         (CONST_VECTOR_STEPPED_P): Likewise.
29204         (CONST_VECTOR_ENCODED_ELT): Likewise.
29205         (const_vec_duplicate_p): Check for a duplicated vector encoding.
29206         (unwrap_const_vec_duplicate): Likewise.
29207         (const_vec_series_p): Check for a non-duplicated vector encoding.
29208         Say that the function only returns true for integer vectors.
29209         * emit-rtl.c: Include rtx-vector-builder.h.
29210         (gen_const_vec_duplicate_1): Delete.
29211         (gen_const_vector): Call gen_const_vec_duplicate instead of
29212         gen_const_vec_duplicate_1.
29213         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
29214         (gen_const_vec_duplicate): Use rtx_vector_builder.
29215         (gen_const_vec_series): Likewise.
29216         (gen_rtx_CONST_VECTOR): Likewise.
29217         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
29218         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
29219         Build a new vector rather than modifying a CONST_VECTOR in-place.
29220         (handle_special_swappables): Update call accordingly.
29221         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
29222         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
29223         Build a new vector rather than modifying a CONST_VECTOR in-place.
29224         (handle_special_swappables): Update call accordingly.
29226 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29228         * simplify-rtx.c (simplify_const_binary_operation): Use
29229         CONST_VECTOR_ELT instead of XVECEXP.
29231 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29233         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
29234         the selector elements to be different from the data elements
29235         if the selector is a VECTOR_CST.
29236         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
29237         ssizetype for the selector.
29239 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29241         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
29242         before testing each element individually.
29243         * tree-vect-generic.c (lower_vec_perm): Likewise.
29245 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29247         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
29248         * selftest-run-tests.c (selftest::run_tests): Call it.
29249         * vector-builder.h (vector_builder::operator ==): New function.
29250         (vector_builder::operator !=): Likewise.
29251         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
29252         (vec_perm_indices::all_from_input_p): New function.
29253         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
29254         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
29255         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
29256         instead of reading the VECTOR_CST directly.  Detect whether both
29257         vector inputs are the same before constructing the vec_perm_indices,
29258         and update the number of inputs argument accordingly.  Use the
29259         utility functions added above.  Only construct sel2 if we need to.
29261 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29263         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
29264         the broadcast of the low byte.
29265         (expand_mult_highpart): Use an explicit encoding for the permutes.
29266         * optabs-query.c (can_mult_highpart_p): Likewise.
29267         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
29268         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29269         (vectorizable_bswap): Likewise.
29270         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
29271         explicit encoding for the power-of-2 permutes.
29272         (vect_permute_store_chain): Likewise.
29273         (vect_grouped_load_supported): Likewise.
29274         (vect_permute_load_chain): Likewise.
29276 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29278         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
29279         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
29280         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
29281         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
29282         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
29283         (vect_gen_perm_mask_any): Likewise.
29285 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29287         * int-vector-builder.h: New file.
29288         * vec-perm-indices.h: Include int-vector-builder.h.
29289         (vec_perm_indices): Redefine as an int_vector_builder.
29290         (auto_vec_perm_indices): Delete.
29291         (vec_perm_builder): Redefine as a stand-alone class.
29292         (vec_perm_indices::vec_perm_indices): New function.
29293         (vec_perm_indices::clamp): Likewise.
29294         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
29295         (vec_perm_indices::new_vector): New function.
29296         (vec_perm_indices::new_expanded_vector): Update for new
29297         vec_perm_indices class.
29298         (vec_perm_indices::rotate_inputs): New function.
29299         (vec_perm_indices::all_in_range_p): Operate directly on the
29300         encoded form, without computing elided elements.
29301         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
29302         encoding.  Update for new vec_perm_indices class.
29303         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
29304         the given vec_perm_builder.
29305         (expand_vec_perm_var): Update vec_perm_builder constructor.
29306         (expand_mult_highpart): Use vec_perm_builder instead of
29307         auto_vec_perm_indices.
29308         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
29309         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
29310         or double series encoding as appropriate.
29311         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
29312         vec_perm_indices instead of auto_vec_perm_indices.
29313         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29314         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
29315         (vect_permute_store_chain): Likewise.
29316         (vect_grouped_load_supported): Likewise.
29317         (vect_permute_load_chain): Likewise.
29318         (vect_shift_permute_load_chain): Likewise.
29319         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
29320         (vect_transform_slp_perm_load): Likewise.
29321         (vect_schedule_slp_instance): Likewise.
29322         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29323         (vectorizable_mask_load_store): Likewise.
29324         (vectorizable_bswap): Likewise.
29325         (vectorizable_store): Likewise.
29326         (vectorizable_load): Likewise.
29327         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
29328         vec_perm_indices instead of auto_vec_perm_indices.  Use
29329         tree_to_vec_perm_builder to read the vector from a tree.
29330         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
29331         vec_perm_builder instead of a vec_perm_indices.
29332         (have_whole_vector_shift): Use vec_perm_builder and
29333         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
29334         truncation to calc_vec_perm_mask_for_shift.
29335         (vect_create_epilog_for_reduction): Likewise.
29336         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
29337         from auto_vec_perm_indices to vec_perm_indices.
29338         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
29339         instead of changing individual elements.
29340         (aarch64_vectorize_vec_perm_const): Use new_vector to install
29341         the vector in d.perm.
29342         * config/arm/arm.c (expand_vec_perm_d::perm): Change
29343         from auto_vec_perm_indices to vec_perm_indices.
29344         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
29345         instead of changing individual elements.
29346         (arm_vectorize_vec_perm_const): Use new_vector to install
29347         the vector in d.perm.
29348         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
29349         Update vec_perm_builder constructor.
29350         (rs6000_expand_interleave): Likewise.
29351         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
29352         (rs6000_expand_interleave): Likewise.
29354 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29356         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
29357         to qimode could truncate the indices.
29358         * optabs.c (expand_vec_perm_var): Likewise.
29360 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29362         * Makefile.in (OBJS): Add vec-perm-indices.o.
29363         * vec-perm-indices.h: New file.
29364         * vec-perm-indices.c: Likewise.
29365         * target.h (vec_perm_indices): Replace with a forward class
29366         declaration.
29367         (auto_vec_perm_indices): Move to vec-perm-indices.h.
29368         * optabs.h: Include vec-perm-indices.h.
29369         (expand_vec_perm): Delete.
29370         (selector_fits_mode_p, expand_vec_perm_var): Declare.
29371         (expand_vec_perm_const): Declare.
29372         * target.def (vec_perm_const_ok): Replace with...
29373         (vec_perm_const): ...this new hook.
29374         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
29375         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
29376         * doc/tm.texi: Regenerate.
29377         * optabs.def (vec_perm_const): Delete.
29378         * doc/md.texi (vec_perm_const): Likewise.
29379         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
29380         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
29381         expand_vec_perm for constant permutation vectors.  Assert that
29382         the mode of variable permutation vectors is the integer equivalent
29383         of the mode that is being permuted.
29384         * optabs-query.h (selector_fits_mode_p): Declare.
29385         * optabs-query.c: Include vec-perm-indices.h.
29386         (selector_fits_mode_p): New function.
29387         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
29388         is defined, instead of checking whether the vec_perm_const_optab
29389         exists.  Use targetm.vectorize.vec_perm_const instead of
29390         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
29391         fit in the vector mode before using a variable permute.
29392         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
29393         vec_perm_indices instead of an rtx.
29394         (expand_vec_perm): Replace with...
29395         (expand_vec_perm_const): ...this new function.  Take the selector
29396         as a vec_perm_indices rather than an rtx.  Also take the mode of
29397         the selector.  Update call to shift_amt_for_vec_perm_mask.
29398         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
29399         Use vec_perm_indices::new_expanded_vector to expand the original
29400         selector into bytes.  Check whether the indices fit in the vector
29401         mode before using a variable permute.
29402         (expand_vec_perm_var): Make global.
29403         (expand_mult_highpart): Use expand_vec_perm_const.
29404         * fold-const.c: Includes vec-perm-indices.h.
29405         * tree-ssa-forwprop.c: Likewise.
29406         * tree-vect-data-refs.c: Likewise.
29407         * tree-vect-generic.c: Likewise.
29408         * tree-vect-loop.c: Likewise.
29409         * tree-vect-slp.c: Likewise.
29410         * tree-vect-stmts.c: Likewise.
29411         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
29412         Delete.
29413         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
29414         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
29415         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
29416         (aarch64_vectorize_vec_perm_const): ...this new function.
29417         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
29418         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29419         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
29420         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
29421         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
29422         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29423         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
29424         into...
29425         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
29426         check for NEON modes.
29427         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
29428         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
29429         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
29430         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
29431         into...
29432         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
29433         the old VEC_PERM_CONST conditions.
29434         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
29435         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
29436         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
29437         (ia64_vectorize_vec_perm_const_ok): Merge into...
29438         (ia64_vectorize_vec_perm_const): ...this new function.
29439         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
29440         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
29441         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
29442         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
29443         * config/mips/mips.c (mips_expand_vec_perm_const)
29444         (mips_vectorize_vec_perm_const_ok): Merge into...
29445         (mips_vectorize_vec_perm_const): ...this new function.
29446         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
29447         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
29448         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
29449         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
29450         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
29451         (rs6000_expand_vec_perm_const): Delete.
29452         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
29453         Delete.
29454         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29455         (altivec_expand_vec_perm_const_le): Take each operand individually.
29456         Operate on constant selectors rather than rtxes.
29457         (altivec_expand_vec_perm_const): Likewise.  Update call to
29458         altivec_expand_vec_perm_const_le.
29459         (rs6000_expand_vec_perm_const): Delete.
29460         (rs6000_vectorize_vec_perm_const_ok): Delete.
29461         (rs6000_vectorize_vec_perm_const): New function.
29462         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
29463         an element count and rtx array.
29464         (rs6000_expand_extract_even): Update call accordingly.
29465         (rs6000_expand_interleave): Likewise.
29466         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
29467         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
29468         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
29469         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
29470         (rs6000_expand_vec_perm_const): Delete.
29471         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
29472         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29473         (altivec_expand_vec_perm_const_le): Take each operand individually.
29474         Operate on constant selectors rather than rtxes.
29475         (altivec_expand_vec_perm_const): Likewise.  Update call to
29476         altivec_expand_vec_perm_const_le.
29477         (rs6000_expand_vec_perm_const): Delete.
29478         (rs6000_vectorize_vec_perm_const_ok): Delete.
29479         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
29480         reference to the SPE evmerge intructions.
29481         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
29482         an element count and rtx array.
29483         (rs6000_expand_extract_even): Update call accordingly.
29484         (rs6000_expand_interleave): Likewise.
29485         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
29486         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
29487         new function.
29488         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
29490 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29492         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
29493         vector mode and that that mode matches the mode of the data
29494         being permuted.
29495         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
29496         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
29497         directly using expand_vec_perm_1 when forcing selectors into
29498         registers.
29499         (expand_vec_perm_var): New function, split out from expand_vec_perm.
29501 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29503         * optabs-query.h (can_vec_perm_p): Delete.
29504         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
29505         * optabs-query.c (can_vec_perm_p): Split into...
29506         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
29507         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
29508         particular selector is valid.
29509         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
29510         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
29511         (vect_grouped_load_supported): Likewise.
29512         (vect_shift_permute_load_chain): Likewise.
29513         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
29514         (vect_transform_slp_perm_load): Likewise.
29515         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
29516         (vectorizable_bswap): Likewise.
29517         (vect_gen_perm_mask_checked): Likewise.
29518         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
29519         implementations of variable permutation vectors into account
29520         when deciding which selector to use.
29521         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
29522         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
29523         with a false third argument.
29524         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
29525         to test whether the constant selector is valid and can_vec_perm_var_p
29526         to test whether a variable selector is valid.
29528 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29530         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
29531         * optabs-query.c (can_vec_perm_p): Likewise.
29532         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
29533         instead of vec_perm_indices.
29534         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
29535         (vect_gen_perm_mask_checked): Likewise,
29536         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
29537         (vect_gen_perm_mask_checked): Likewise,
29539 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
29541         * optabs-query.h (qimode_for_vec_perm): Declare.
29542         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
29543         (qimode_for_vec_perm): ...this new function.
29544         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
29546 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
29548         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
29549         does not have a conditional at the top.
29551 2018-01-02  Richard Biener  <rguenther@suse.de>
29553         * ipa-inline.c (big_speedup_p): Fix expression.
29555 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
29557         PR target/81616
29558         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
29559         for generic 4->6.
29561 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
29563         PR target/81616
29564         Generic tuning.
29565         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
29566         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
29567         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
29568         cond_taken_branch_cost 3->4.
29570 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
29572         PR tree-optimization/83581
29573         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
29574         TODO_cleanup_cfg if any changes have been made.
29576         PR middle-end/83608
29577         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
29578         convert_modes if target mode has the right side, but different mode
29579         class.
29581         PR middle-end/83609
29582         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
29583         last argument when extracting from CONCAT.  If either from_real or
29584         from_imag is NULL, use expansion through memory.  If result is not
29585         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
29586         the parts directly to inner mode, if even that fails, use expansion
29587         through memory.
29589         PR middle-end/83623
29590         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
29591         check for bswap in mode rather than HImode and use that in expand_unop
29592         too.
29594 Copyright (C) 2018 Free Software Foundation, Inc.
29596 Copying and distribution of this file, with or without modification,
29597 are permitted in any medium without royalty provided the copyright
29598 notice and this notice are preserved.